Oyster larvae (Crassostrea virginica) could enhance their settlement success by moving toward the seafloor in the strong turbulence associated with coastal habitats. We characterized the behavior of individual oyster larvae in grid-generated turbulence by measuring larval velocities and flow velocities simultaneously using infrared particle image velocimetry. We estimated larval behavioral velocities and propulsive forces as functions of the kinetic energy dissipation rate ε, strain rate γ, vorticity ξ and acceleration α. In calm water most larvae had near-zero vertical velocities despite propelling themselves upward (swimming). In stronger turbulence all larvae used more propulsive force, but relative to the larval axis, larvae propelled themselves downward (diving) instead of upward more frequently and more forcefully. Vertical velocity magnitudes of both swimmers and divers increased with turbulence, but the swimming velocity leveled off as larvae were rotated away from their stable, velum-up orientation in strong turbulence. Diving speeds rose steadily with turbulence intensity to several times the terminal fall velocity in still water. Rapid dives may require a switch from ciliary swimming to another propulsive mode such as flapping the velum, which would become energetically efficient at the intermediate Reynolds numbers attained by larvae in strong turbulence. We expected larvae to respond to spatial or temporal velocity gradients, but although the diving frequency changed abruptly at a threshold acceleration, the variation in propulsive force and behavioral velocity was best explained by the dissipation rate. Downward propulsion could enhance oyster larval settlement by raising the probability of larval contact with oyster reef patches.
Many mollusc veligers change their behavior in response to turbulence, and variations in these responses provide clues to whether or how adult habitat structure shapes larval behavior. Veligers pull in the velum and sink when disturbed (Barile et al., 1994; Young, 1995), but reactions to turbulence vary among species from enclosed habitats versus exposed coastlines. For example, mud snails (Ilyanassa obsoleta) and blue mussels (Mytilus edulis) inhabit estuaries or inlets with energetic tidal currents, and larvae of these species swim up in calm water but sink in turbulence above a threshold value of the turbulent kinetic energy dissipation rate ε (Fuchs et al., 2004; Fuchs and DiBacco, 2011). These behaviors could raise the probability of being retained near and settling in turbulent coastal inlets (Fuchs et al., 2007; Fuchs and DiBacco, 2011). Snail larvae (Crepidula spp. and Anachis spp.) from subtidal beaches behave differently, sinking in calm water and swimming up in strong turbulence (Fuchs et al., 2010). These genus-specific responses to turbulence suggest that larval behaviors may be adapted for settlement into distinct adult habitat types. Here, we studied larval responses to turbulence in a reef-building bivalve, the eastern oyster Crassostrea virginica. Oyster larvae undergo rapid downward accelerations, termed ‘dive bombing’ (Finelli and Wethey, 2003), but the trigger for diving is unknown. We hypothesize that oyster larvae dive in response to turbulence as a way of concentrating near the bottom despite vigorous mixing over the rough substrates created by oyster reefs.
It is generally assumed that veliger larvae descend in turbulence by passive gravitational sinking. Passive descents are a reasonable assumption because veligers are negatively buoyant and sink by arresting the ciliary beat or retracting the velum when disturbed or presented with chemical cues in still water (e.g. Fretter, 1967; Cragg, 1980; Hadfield and Koehl, 2004). Water motion makes it more difficult to observe the cilia and velum, however, and active descents cannot be ruled out without estimates of propulsive force under realistic flow conditions. The downward accelerations observed in oyster larvae (Finelli and Wethey, 2003) could indicate an abrupt behavioral change from upward swimming to passive sinking, an abrupt reduction in upward propulsive force, or a change in the direction of propulsion. Propulsive force can be estimated from measured velocities of larvae and the flow around them. Such measurements are difficult, and previous studies on larvae in turbulence described larval behavior only in terms of behavioral velocities (Fuchs et al., 2004; Fuchs et al., 2010; Fuchs and DiBacco, 2011).
Here, we describe both larval behavioral velocity and propulsive forces in turbulence, and this added complexity calls for a definition of terms to distinguish among modes of behavior. We use ‘ascent’ or ‘descent’ to refer to a positive or negative vertical velocity due to larval propulsion. These behavioral velocities are exclusive of fluid motions and are distinct from the net larval velocity due to the combined motions of larvae and fluid. We define ‘swimming’ and ‘diving’ as propulsive forces directed upward and downward, respectively, relative to the larval axis. By this definition, swimmers can ascend or descend, depending on the magnitude of propulsive force relative to the combined opposing forces of drag and gravity. We use ‘sinking’ to refer only to those larvae descending passively without propulsion.
Previous studies described larval responses to turbulence only as a population- and time-averaged function of the dissipation rate. The dissipation rate is a good descriptor of larval-scale turbulence because it defines the Kolmogorov length, time and velocity scales: ηk=(ν3/ε)0.25, τk=(ν/ε)0.5 and υk=(νε)0.25, respectively, where ν is kinematic viscosity (see List of symbols). These scales represent the smallest eddies with which larvae may interact. Larvae respond rapidly to instantaneous cues (e.g. Hadfield and Koehl, 2004; Koehl and Hadfield, 2010), so although population-average behaviors are useful for modeling settlement processes (Fuchs et al., 2007), average behaviors may be unrepresentative of larval reactions to instantaneous turbulence. Moreover, we suspect that larvae cannot detect the dissipation rate itself but rather sense and respond to more specific flow characteristics such as the strain rate γ (deformational shear), vorticity ξ (rotational shear) or acceleration α (e.g. Kiørboe et al., 1999). These velocity gradients probably elicit behavioral changes when the magnitudes of γ, ξ or α exceed the larval detection limits or response thresholds. If larvae react to instantaneous velocity gradients, then time-resolved observations are needed to characterize responses to turbulence.
Larvae potentially sense turbulence with the velar cilia, used for swimming and feeding, or with statocysts, used to detect gravity (Chia et al., 1981). Some veligers have mechanosensory cilia that stop beating or draw inwards when touched (Murakami and Takahashi, 1975; Mackie et al., 1976; Dickinson, 2002). Deformation of the cilia could enable larvae to sense strain rates, or the whole ciliated velum could act as an antenna to detect spatial variability in the shear. Statocysts could sense changes in orientation (vorticity-induced rotation) or changes in velocity (acceleration). Veligers have an asymmetric density distribution and normally swim with the velum facing up, but they can rotate away from this passively stable orientation when the viscous torque due to vorticity or shear across the body exceeds the gravitational torque (Kessler, 1986; Jonsson et al., 1991). Vorticity and acceleration would likely be sensed only with the statocyst, whereas strain rate may be detectable both by the cilia as deformation and by the statocysts as axial rotation. Pinpointing the sensing mechanism will require an understanding of which velocity gradients elicit changes in behavior.
We investigated the behavioral responses of oyster larvae to dissipation rates and velocity gradients. Larval velocities and water velocities were measured simultaneously using infrared particle-image velocimetry (IR PIV) (e.g. Catton et al., 2007; Sutherland et al., 2011), and larval propulsive forces were estimated using an expanded equation of particle motion. These detailed measurements enabled us to characterize the velocities and propulsive forces of individual larvae as a response to instantaneous flow characteristics. This combined study of behavioral velocities and propulsive forces adds a new dimension to our insights into how larvae respond to turbulence.
MATERIALS AND METHODS
Behaviors of eyed oyster larvae (C. virginica Gmelin 1791) were characterized in both still water and turbulence. We measured water velocities and larval velocities simultaneously using near-IR PIV. This method requires seeding the flow with particles, illuminating a plane with a laser light sheet, and taking pairs of images separated by a small time step. The image pairs are used to calculate water velocities in the plane based on the motions of seeding particles within small interrogation areas (Adrian, 1991). Seeding particles could alter larval behavior, so in summer 2010 we characterized behavior of larvae in still water with different particle types. In April 2011 we carried out turbulence experiments in a grid-stirred tank.
Larvae were shipped overnight from Horn Point Laboratory and used within 48 h. Before use, larvae were kept in 10 l cultures at 20°C and a salinity of 9.5 SP with Shellfish Diet (Reed Mariculture, Campbell, CA, USA) mixed algae for food. All experiments were done at room temperature (21–22°C) and a salinity of 9.5 SP.
For still-water experiments, larvae were added to 8 liter aquaria containing no particles (control) or one of three different particle types: mixed algae (3–18 μm, ~1.07 g cm−3, Shellfish Diet), hollow glass spheres (12 μm, ~1.1 g cm−3, Sphericel, Potters Industries LLC, Valley Forge, PA, USA) or nylon particles (20 μm, ~1.03 g cm−3, PSP, Dantec Dynamics Inc., Holtsville, NY, USA). Larval and particle concentrations were 0.3–0.7 larvae ml−1 and 5.0×104 cells ml−1, respectively. The particle concentrations were comparable to typical feeding concentrations for larval cultures and the concentrations of seeding particles required for PIV. No-particle controls were replicated nine times, and particle treatments were replicated six times. For each treatment we used an infared LED spotlight to illuminate the aquarium and video-recorded larval motions for 12–15 min at a frame rate of 3 Hz using a digital video camera (KPF-120, Hitachi Ltd, Tokyo, Japan) and capture software (XCAP, EPIX Inc., Buffalo Grove, IL, USA). We later reconstructed the larval trajectories (N=18–3498 per replicate) using a custom particle-tracking algorithm in Matlab (e.g. Fuchs et al., 2004) to estimate larval velocities.
Larvae were subsampled after each replicate for measurement of shell length and terminal sinking velocity. Shell length (N=30–35 per replicate) was measured digitally using a stereomicroscope and software (M205C and Leica Application Suite, Leica, Wetzlar, Germany). Fall velocity (N=47–124 per replicate) was measured from digital video of ethanol-killed larvae sinking through a 2 liter settling column at room temperature and a salinity of 9.5 SP.
Turbulence experiments were done in a 170 liter tank (46 cm wide × 46 cm deep × 80 cm high) with turbulence generated by vertical oscillation of two horizontal stirring grids. The grids had a mesh size of 6.35 cm, a grid separation distance of 40.6 cm and an oscillation amplitude of 12.7 cm. Six different stirring frequencies were used, ranging from f=0.02 to 1.61 Hz. Unlike tanks with a single stirring grid (e.g. Hopfinger and Toly, 1976; Brumley and Jirka, 1987), tanks with two stirring grids produce turbulence that is homogeneous and nearly isotropic in a large region centered between the two grids (Srdic et al., 1996; Shy et al., 1997).
Measurements were made with an IR PIV system that included a pulsed diode laser (NanoPower 7 W, 808 nm) with a ~2 mm beam width and a 4 megapixel camera (FlowSense, Dantec Dynamics) with a 55 mm lens (Leica). We used ~18 μm concentrated algae (Thalassiosira weissflogii, Reed Mariculture) as seeding particles because artificial particles induced behavioral changes in still-water experiments. A foam lid was used to dampen secondary flows. The PIV image plane (5 cm high × 10 cm wide) was centered at z=20.3 cm from each grid and 13 cm from each of the nearest walls, an offset of 10 cm from the center. The horizontal offset was necessary because the IR laser light attenuated with distance from the source and was too weak in the center of the tank. The images were far enough from the walls that larval motions were free of wall effects (Vogel, 1994).
Two replicates were done using larval concentrations of 0.5 and 0.3 larvae ml−1, respectively. For the first replicate we used six, randomly ordered turbulence levels. A 10 min warm-up period at the beginning of each treatment ensured that the turbulence was stationary. For the second replicate we used a different treatment order but were only able to complete three turbulence treatments because of an equipment malfunction. At each turbulence level we collected 10 min of PIV data at 10 Hz (i.e. 10 image pairs per second), observing hundreds to thousands of individual larvae per treatment. All data were combined in our analysis.
The PIV images of larvae in turbulence represent a two-phase flow, with larvae and fluid moving in different directions, so we separated the images of larvae and tracer particles (e.g. Kiger and Pan, 2000) to quantify larval and fluid motions. Before calculating the fluid velocity vectors we equalized the image backgrounds, removed noise and masked out the larvae to obtain good estimates of background fluid flow and to limit error in the calculation of individual larval velocities. The image intensity varied spatially because of IR light attenuation, so we were unable to use standard procedures of subtracting the mean background intensity of each individual image and removing pixel-scale noise with a median filter (Khalitov and Longmire, 2002; Cheng et al., 2010). Instead, we equalized the background by calculating the mean image intensity over each 10 min sampling interval and subtracting the mean intensity from each image, repeating for frame 1 and frame 2 images. The particle image intensity also varied spatially, so we used wavelet analysis (e.g. Torrence and Compo, 1998; Weng et al., 2001) to remove the noise based on its spatial scale while ignoring spatial variability in particle image intensity. To reduce noise in the images we decomposed each image using Coiflet wavelets (Mohideen et al., 2008), removed wavelet coefficients below a scale threshold and reconstructed the image from the remaining signal. The resulting images had a relatively constant background intensity, were free of small-scale noise, and retained the scale and intensity of the particle images.
We also had to remove larvae from the images, because larval velocities often exceeded or opposed the underlying flow velocities. Larval particle images sometimes became saturated and had a bright, reflective halo, so we first applied a 2-dimensional, high-pass, fast Fourier transform filter that reduced the halo effect. After filtering, we removed the residual background by squaring the image intensity and setting to zero any pixel intensities below a threshold. Lastly, we binarized the images, identified and labeled each particle, and classified particles with area >10 pixels as larvae. Larval particle images were removed, leaving images of only seeding particles.
Fluid velocities and turbulence
The paired images of seeding particles were processed using adaptive correlation algorithms in Dynamic Studio (Dantec) to calculate velocity vectors u and w in the x and z directions, respectively. We used interrogation areas of 64×64 pixels at the two lowest settings and 32×32 pixels at higher settings with a 50% overlap to give vector resolutions of Δx=0.16 cm and Δx=0.08 cm, respectively. These resolutions gave the best balance between improving the quality of vector calculations and limiting the difference between the vector spacing Δx and Kolmogorov length scale ηk.
Larval behavioral velocities
Larval velocities were calculated by reconstructing larval trajectories from the original images. Larvae were much larger than the algal seeding particles and were easily classified based on their equivalent spherical diameter, solidity and eccentricity. We analyzed only larvae with area >20 pixels that could be tracked unequivocally between paired frames and from image pair to image pair. Paired frames 1 and 2 were treated as frames of two separate image sequences. We reconstructed the larval trajectories in each sequence by particle tracking in Matlab and then matched larvae in the two sequences to get paired trajectories offset by δt, the time between paired frames. Larvae with trajectories in only one sequence or with trajectories of unequal lengths in the two sequences were excluded from the analysis. We analyzed the paired trajectories of 6355 larvae, including 40,268 instantaneous observations. Trajectory durations ranged from 0.89±1.21 s (mean ± 1 s.d.) at the lowest turbulence level to 0.29±0.16 s at the highest turbulence level.
To limit velocity errors, we estimated larval velocities from the sequence trajectories rather than from movements between paired frames. Particle displacements have uncertainty due to errors in the calculation of particle centroid positions (±0.1–0.25 pixels) (Wernet and Pline, 1993; Adrian, 1997), but these uncertainties can be offset by using a longer time step to increase the dynamic velocity range (Adrian, 1997). The larvae had an average image diameter of 10.8 pixels, or about 549 μm, and the time between image pairs was Δt=0.1 s, giving a velocity-error standard deviation of 1.6×10−3 cm s−1 [eqns 2,3 in Adrian (Adrian, 1997)].
Although larval velocities were calculated from sequence trajectories, water velocities and flow statistics were calculated by PIV from each image pair. To characterize the instantaneous flow corresponding to each larva's velocity, we interpolated the water velocities, u and w, and the turbulence characteristics, α, γ, ξ and ε, to the larval positions at each time step. We used an unweighted linear interpolation because it gave results nearly identical to those from a more accurate spline interpolation and required less computation time. The interpolated water velocities and turbulence characteristics were then averaged for each larval trajectory segment.
We estimated larval behavioral velocities as ub=uo–u and wb=wo–w, where uo and wo are the observed horizontal and vertical velocities from larval trajectories and u and w are instantaneous fluid velocities interpolated to larval positions. Behavioral vertical velocities wb are a vector sum of the vertical velocity the larva generates by propulsion and the gravitational sinking velocity. These estimates require the assumption that larval velocities and fluid velocities are additive (e.g. Reeks, 1977). Because larvae are denser than seawater, however, they will have some additional ‘slip’ velocity when the water accelerates (e.g. Maxey and Riley, 1983; Kiørboe and Visser, 1999). Here, we were unable to separate the behavioral velocity from the slip velocity because flow was unsteady. We estimated the maximum slip velocities for individual larvae assuming steady-state acceleration [eqns 12–15 in Kiørboe and Visser (Kiørboe and Visser, 1999)] and found that the average slip velocity was <1% of the estimated behavioral velocity. Given that the slip velocity was small compared with behavioral velocity, the omission of slip velocity contributes negligible uncertainty to our analysis.
We used paired trajectories to calculate the larval Lagrangian accelerations. The net larval acceleration is , where is the observed translational velocity, is the behavioral component of the larval translational velocity, is the fluid velocity at the larva's location, an over-arrow denotes a vector, and the subscripts 1 and 2 indicate the sequence number. Here , and are 2-dimensional projections of 3-dimensional motion. Sequence 2 was used only for calculating larval accelerations. All other calculations were based on sequence 1 trajectories, and sequence subscripts are omitted hereafter.
Force balance and terminal velocity
Larval behavior in still water
Still-water experiments confirmed that the presence of artificial particles altered larval behavior (Table 1). Larvae in controls and algal treatments had similar average density, terminal velocity, swimming velocity, direction of motion, sinking frequency and propulsive force. Larval density estimates (ρp=1.15±0.02 g cm−3) were in the range reported previously for bivalve veligers (ρp=1.1–1.22 g cm−3) (Jonsson et al., 1991; Finelli and Wethey, 2003; Schwalb and Ackerman, 2011). Larval propulsion was directed upward in 98% of the larvae, yet the average swimming velocities were near zero and slightly negative, indicating a mix of ascending and descending swimmers. Larvae in the glass and nylon particle treatments had more positive swimming velocities and used more propulsive force than those in controls or algal treatments. Most notably, far fewer larvae were observed in the artificial particle treatments than in controls or algae treatments. The number of tracks, normalized by the number of larvae, video recording time, and image area, was an order of magnitude lower in the glass and nylon particle treatments than in the control and algae treatments. This result supports our qualitative observation that when exposed to glass or nylon particles, many larvae sank immediately to the bottom and remained there, suggesting an adverse reaction to artificial particles.
Turbulence treatments spanned a wide range of turbulence conditions, with fluid Reynolds numbers ranging from Re=36 to 860 (Table 2), where Re=VRMSℓ/ν, V=(2u2+w2)0.5, the subscript RMS indicates a root mean square, ℓ=0.2zo, and ℓ is the eddy length scale at a distance zo from the grids. The spatially averaged dissipation rates were ε=4.5×10−4 to 4.0 cm2 s−3, with corresponding Kolmogorov length scales of ηk=0.22–0.02 cm. The characteristic eddy length scale can be estimated by the Taylor microscale λ=(15νVRMS2/ε)0.5 and ranged from 0.43 to 1.66 cm. To obtain highly accurate dissipation rate estimates, the vector resolution Δx should be close to the Kolmogorov length scale (1≤Δx/η<3) and less than 30% of the Taylor microscale λ (Antonia et al., 1994; Saarenrinne and Piirto, 2000; Tanaka and Eaton, 2007; de Jong et al., 2009). Here, Δx/η ranged from 0.73 to 4.0 and Δx was 10–19% of λ. Based on the Δx/η criterion, ε may have been underestimated by up to ~10% at the highest turbulence level (Antonia et al., 1994). These errors are negligible for the behavior analysis given that measured dissipation rates spanned four orders of magnitude. Mean flows were upward, and turbulence was relatively anisotropic with isotropy ratios of wRMS/uRMS=1.41–1.64. This deviation from isotropy indicates the presence of weak secondary flows and was an unavoidable consequence of making measurements away from the center of the tank.
Larval behavior in turbulence
The apparent range of larval vertical behavioral velocities varied with the turbulence characteristic used for binning (Fig. 2). Average larval velocities wb were always slightly above zero in weaker turbulence and then became increasingly negative in turbulence above a threshold level. The average descent speed was about one-third higher when larval velocity was binned by acceleration or dissipation rate than when it was binned by strain rate or vorticity, indicating that rapid descents were most strongly associated with high accelerations and high dissipation rates. Based on the range of average velocities, the variation in wb was best explained by dissipation rate, followed by acceleration, vorticity and strain rate. The fitted behavior model (Eqn 5) also gave the highest coefficient of determination for dissipation rate, followed by acceleration, vorticity and strain rate (Table 3). Estimates from Eqn 5 indicate that larval velocities switched from positive to negative at threshold values of αcr=3.78×10−1 cm s−2, γcr=1.34×10−1 s−1, ξcr=3.64×10−1 s−1 and εcr=7.78×10−2 cm2 s−3.
The force balance analysis demonstrated patterns in the direction and magnitude of propulsive force as a response to turbulence. The fraction of larvae propelling themselves downward was strongly dependent on turbulence (Fig. 3). The fractions of swimmers and divers changed most abruptly when larvae were averaged in small bins of acceleration, with a sudden change of slope corresponding to the threshold acceleration αcr. The fractions of swimmers and divers changed more gradually when larvae were binned by strain rate, vorticity or dissipation rate. The classification of larvae as swimmers or divers was generally insensitive to L. Although the rotation angle φ varied widely with L, the propulsion angle θV varied little (Fig. 4), and larvae rarely experienced both a large φ and a large θV simultaneously.
All larvae experienced a larger average rotation angle φ in stronger turbulence because of increasing vorticity, but the average direction of propulsion θV relative to the larval axis remained steady (Fig. 4). For divers, φ at low dissipation rates (Fig. 4A) was generally smaller and less variable than φ at low accelerations (Fig. 4B) or strain rates (not shown). This inconsistency may arise because the rotation angle is defined by vorticity, which is more strongly correlated with dissipation rate than with acceleration or strain rate. In larval coordinates, the propulsive force was consistently directed at an average angle of θV≈90 deg for swimmers and θV≈−90 deg for divers (Fig. 4C,D), although θV for divers was variable in weak turbulence where diving was infrequent.
Both swimmers and divers used more propulsive force and had higher behavioral velocity magnitudes |wb| in stronger turbulence (Fig. 5). Swimmers directed their propulsive force upward and showed a steady rise in with turbulence, so their vertical velocities increased from near zero in calm water to wb≈0.5 cm s−1 in intermediate turbulence. Despite the steady rise in propulsive force, swimmers' velocities leveled off and even dropped in strong turbulence, presumably because larvae rotated and their propulsive force was directed away from vertical. Diving larvae had a more complex response to turbulence. In weaker turbulence, and wb grew steadily with turbulence when larvae were binned by ε but were extremely variable when larvae were binned by α, γ and ξ. At low ε the diving velocities were near the estimated terminal fall velocity (ws=−0.58±0.11 cm s−1). In stronger turbulence, and |wb| of divers grew steadily with turbulence regardless of which characteristic was used for binning. At the highest ε the diving velocities reached wb≈−3 cm s−1, five times the terminal fall velocity of passive larvae. The variation in propulsive force and velocity of larval dives was best explained by dissipation rate, particularly in weaker turbulence.
The apparent dependence of behavioral velocity and propulsive force on dissipation rate was further supported by the relationships between or wb and the rotation angle φ (Fig. 6). When larvae were grouped in small bins of dissipation rate, the propulsive forces and diving velocities were highly correlated with the larval rotation angle (R2≥0.93 for linear regressions). For swimmers, the relationship between wb and φ appeared more non-linear because at large rotation angles (|φ|>10 deg) the larval propulsive force was directed away from the positive z direction. The relationships between wb or and φ were weaker when larvae were binned by acceleration or strain rate and weakest when larvae were binned by vorticity ξ, even though rotation angle was estimated directly from vorticity. This result implies that the strength of a diving reaction depends less on axial rotation than on more general features of small-scale turbulence.
The observed behaviors of oyster larvae provide intriguing new insights into how and why larvae respond to turbulence. The use of IR PIV enabled us to measure simultaneously the behavioral velocities and propulsive forces of individual larvae as a response to instantaneous turbulence. Like other veligers (Fuchs et al., 2004; Fuchs and DiBacco, 2011), oyster larvae frequently descended in strong turbulence. Unlike other veligers, however, oyster larvae reached descent velocities that greatly exceeded the terminal fall velocity of passive larvae (Fig. 7A). Our results suggest that oyster larvae undergo a behavioral shift from infrequent, nearly passive descents in weaker turbulence to frequent, active dives in stronger turbulence. Active diving has little precedent among invertebrate larvae. Diving would require an energy expenditure and is an exciting contrast to previous observations of mollusc larvae that sink passively by retracting the velum. If oysters have developed energetically demanding strategies to achieve high diving speeds, this implies that there is strong selective pressure for larvae to descend in turbulent environments.
Before discussing the implications of active diving, we first address whether the observed descent speeds could arise solely from turbulence-enhanced passive sinking (e.g. Ruiz et al., 2004). Passive larvae could have different average sinking speeds in turbulence than in still water if the larvae have inertia. Inertia is negligible for passive particles at particle Reynolds numbers of Rep<0.5, and at low Rep the average terminal velocity is unaffected by turbulence (Reeks, 1977). At Rep>0.5 particles gain inertia, and the average sinking velocity can be lower or higher in turbulence than in still water. Oyster larvae would have Rep=1.7 when sinking passively at the terminal velocity. The larvae observed in turbulence had Rep up to 8.8 on average and up to 25.7 for individual larvae (Fig. 7B). Given these intermediate particle Reynolds numbers, we must consider whether larval inertia contributed to higher sinking velocities in turbulence.
Scales of larvae and turbulence
For inertial particles, turbulence has the greatest effect on particle velocity when particle velocity is similar to the Kolmogorov velocity scale, ws/υk≈1, when particle size differs from the Kolmogorov length scale, d/ηk≠1, and when the particle response time τp=d2ρp(18νρf)−1 is similar to the Kolmogorov time scale, giving a Stokes number of St=τp/τk≈1 (e.g. Wang and Maxey, 1993). Terminal velocity can also be reduced by added drag on non-spherical shapes, but the effects of shape are small at Rep<10 (Komar and Reimers, 1978; Davies, 1979). For passive sinkers, the terminal fall velocity was comparable to the Kolmogorov velocity scale in the strongest turbulence, with ws/υk≈1 at ε=10 cm2 s−3 (Fig. 7B). For observed larvae, the velocity-scale ratios were always greater than one, with for swimmers and 5.9–17.2 for divers. Thus, by the velocity criterion, turbulence may have sped up the descent of passively sinking larvae, although the observed larvae were less likely to experience the same effect.
By the size and time scale criteria, in contrast, turbulence could have slowed or had no effect on larval descent (Fig. 7B). Inertial particles that are smaller than ηk tend to concentrate in high-strain-rate, low-vorticity regions and can have terminal velocities 27–50% higher in turbulence than in still water (Maxey, 1987; Wang and Maxey, 1993), whereas particles larger than ηk experience more drag and have lower terminal velocities in turbulence than in still water (e.g. Brucato et al., 1998). Oyster larvae were smaller than ηk at low dissipation rates and slightly larger than ηk at ε≥1 cm2 s−3 (Fig. 7B). The ratio of length scales was too close to d/ηk=1 at ε>εcr to expect much effect of turbulence on descent velocities (Wang and Maxey, 1993; Brucato et al., 1998), but by the size criterion the strongest turbulence may have slowed the larval descents. Larval response times were also too short to expect much effect of turbulence on descent velocities. Larval Stokes numbers were generally St<<1, reaching only St≈0.01 at the threshold dissipation rate and St≈0.1 at the highest dissipation rate (Fig. 7B). These low Stokes numbers indicate that larvae had short response times and would be unlikely to form clusters or experience a downward bias in turbulent transport (Salazar et al., 2008).
Based on these considerations of scale, the potential effects of turbulence on descent speeds were inconsistent. Turbulence was unlikely to greatly speed larval descent, and we are confident that the observed descents were active dives rather than turbulence-enhanced passive sinking. Even if we assume that turbulence raised larval descent speeds by the maximum amount (50%) (Wang and Maxey, 1993), the observed descents in strong turbulence could not be explained by passive sinking (Fig. 7A), indicating that larvae actively propelled themselves downward.
Our results provide compelling evidence that larvae dove more frequently and more forcefully in stronger turbulence. Even in still water the propulsive force used for swimming exceeded estimates for other ciliated larvae (=0.5×10−9 to 5.8×10−9 N) (Jonsson et al., 1991; Emlet, 1994; Hansen et al., 2010) because oyster veligers are larger or more dense than larvae studied previously. In turbulence, the propulsive force and behavioral velocity magnitudes of both swimming and diving larvae grew steadily with dissipation rate. A similar turbulence-induced increase in swimming activity was observed in ciliated echinoid blastulae (Dendraster excentricus), which swam faster in stronger shear (McDonald, 2012). Swimming oyster larvae reached ascent velocities of only a few millimeters per second, because although they used more upward propulsive force in stronger turbulence, they also experienced larger axial rotation angles that directed the propulsive force away from vertical. This rotation-induced limitation of larval swimming abilities is consistent with model predictions for larvae that lose their passively stable orientation in shearing flow (Grünbaum and Strathmann, 2003; Clay and Grünbaum, 2010). Diving larvae also experienced large rotation angles in strong turbulence, but they compensated for rotation by using more propulsive force than swimmers and achieved impressive descent speeds of up to a few centimeters per second.
The observed behavioral shift from nearly passive sinking to active diving may require a change in propulsive mode. Mollusc veligers propel themselves by beating the velar cilia, with propulsive forces directed upward relative to the larval axis. A faster ciliary beat generates more propulsive force and a higher upward swimming velocity (Arkett et al., 1987; Gallager, 1993). Descent generally requires less energy than ascent because veligers are denser than water and can descend by slowing the ciliary beat, arresting the cilia or drawing the velum inside the shell (Arkett et al., 1987; Gallager, 1993). Some ciliated larvae can reverse their swimming direction by reversing the direction of ciliary beat (e.g. Lacalli and Gilmour, 1990), but we are unaware of any reports of ciliary reversal in veligers. Without a ciliary reversal it is implausible that diving larvae generated the observed propulsive forces or diving speeds by ciliary swimming alone.
Larvae potentially gained some additional downward thrust by another propulsive mechanism such as flapping the lobes of the velum. Flapping is a common swimming mode among planktonic molluscs. Pteropods flap their parapodia and can do so both as larvae before losing the velum or as adults in alternation with ciliary swimming (Bandel and Hemleben, 1995; Childress and Dudley, 2004; Borrell et al., 2005). Some snail veligers have been observed flapping the velum, although observations are limited to larvae with intermediate particle Reynolds numbers of Rep≈4–10 (H.L.F., unpublished observation) (Lebour, 1931; Manríquez and Castilla, 2011). Flapping of appendages can generate positive thrust even at low particle Reynolds numbers and becomes energetically efficient at Rep=5–20 (Walker, 2002; Childress and Dudley, 2004). In this study, some larvae certainly experienced the range of Rep where flapping would become an energetically efficient mode of propulsion.
Responses to velocity gradients
We expected larvae to change their behavior in response to spatial or temporal velocity gradients, but no gradients emerged as a dominant behavioral cue. In our experiments the velocity gradients were correlated with one another, and their effects on behavior could not be completely isolated. Yet, we found little evidence of abrupt behavioral changes at threshold values of any velocity gradients. One exception was the fraction of diving larvae, which underwent a larger change in slope at the threshold acceleration αcr than at γcr, ξcr or εcr (Fig. 3). Stronger dives were also more closely associated with high accelerations than with high spatial gradients. Overall, however, the strength of a dive appeared most related to dissipation rate, and in weak turbulence the strength of a dive showed no relationship with any turbulence characteristic except dissipation rate. These results suggest that larval dives are a complex reaction to multiple aspects of small-scale turbulence.
Based on the larval responses to turbulence, we suspect that the statocysts may be more important than the velar cilia for turbulence detection. Statocysts could detect accelerations when the statolith is accelerated into the mechanosensory cilia lining the statocyst lumen. The threshold value αcr was associated with an increase in diving frequency and may correspond to an acceleration at which the statolith impacts the cilia with enough force to deflect them by a threshold amount. The statocysts could also detect axial rotation due to vorticity as the statoliths rolled onto cilia around the internal surface (e.g. Gallin and Wiederhold, 1977). The rotation angle had no obvious influence on whether larvae swam or dove but did explain most of the variation in propulsive force, particularly by diving larvae. Statocysts probably play a role in both detecting and responding to turbulence.
Whereas statocysts would detect acceleration or rotation of the larval body, cilia could detect spatial gradients such as strain rate in the surrounding fluid. The ability of larvae to detect strain rate depends on the size of the detector, so it is useful to convert strain rate to a signal strength dγ, where d is an appropriate length scale. The threshold strain rate for oyster larvae gives a signal strength of dγcr≈4×10−3 cm s−1 over the length of a larva. This threshold is 10–100 times lower than signal strengths inducing jumps in the most sensitive copepods (0.02 cm s−2) (Kiørboe et al., 1999) and comparable to signal strengths inducing jumps in the most sensitive ciliates and flagellates (3.1×10−3 cm s−2) (Jakobsen, 2001). If we assume a cilium length of dc≤50 μm (e.g. Sleigh and Blake, 1977), the threshold signal strength over a cilium is dcγcr <7×10−4 cm s−1, lower than any threshold previously observed. The estimated threshold signal strengths are improbably low, and the larvae exhibited no abrupt behavioral changes at those thresholds, so it is unlikely that velar cilia are solely responsible for sensing turbulence.
Oyster larvae exhibited an extraordinary diving behavior that would enable them to rapidly approach the seabed. Rapid descents may confer large fitness gains because unlike most shallow-water species, oysters form discrete reefs on intertidal or subtidal mud flats. These reefs are patchy, tens to hundreds of meters long, and rougher than surrounding substrates. Natural oyster reefs have drag coefficients of Cd≈0.11 (Whitman and Reidenbach, 2012), 10–100 times greater than those over flat mud or sand (e.g. Green et al., 1998; Geyer et al., 2000; Whitman and Reidenbach, 2012). Drag coefficients are related to shear velocity u* by the quadratic drag law, u*=Cd0.5 U, and dissipation rate can be estimated as ε=u*3/κz, where κ=0.41 is von Karman's constant. Based on these simple models and observed drag coefficients, the dissipation rates should also be 10–100 times higher over oyster reefs than over surrounding mud flats. Larvae that respond to high dissipation rates by descending would be more likely to concentrate near the bed over a reef than over the flats. Descent speed may be critical in determining whether larvae contact a reef patch before passing over onto flatter substrates. Larvae would have better odds of hitting an oyster reef if they dive actively than if they sink passively, and the improved settlement odds may confer fitness benefits that offset the energetic cost of active downward propulsion. Settlement rates could be further enhanced by responses to chemical cues near the bed (e.g. Turner et al., 1994; Koehl and Reidenbach, 2007). Using a numerical model that will be presented elsewhere, we are investigating how larval behaviors interact with substrate type to affect oyster settlement.
We thank D. Merritt and S. Alexander at Horn Point Laboratory for providing the oyster larvae. G. Gerbi, F. J. Diez, A. Christman, K. Helfrich and L. Mullineaux contributed to helpful discussions, and G. Gerbi, J. P. Grassle and two anonymous reviewers provided constructive comments on the manuscript.
This research was supported by the National Science Foundation (NSF) [grant no. OCE-1060622 to H.L.F.]. E.L.S. was supported by a Research Internship in Ocean Sciences [NSF grant no. OCE-1062894 to G. Taghon].
LIST OF SYMBOLS
- d, r
larval shell length and radius
added mass or acceleration reaction force vector
Basset or Boussinesq force vector
viscous Stokes drag force vector
form drag force vector
pressure gradient force vector
velar propulsion force vector
weight force vector
acceleration due to gravity
distance between centers of buoyancy and gravity
fluid Reynolds number
particle Reynolds number
- u, w
vertical and horizontal fluid velocity
- ub, wb
vertical and horizontal larval behavioral velocity
- uo, wo
vertical and horizontal observed larval velocity
behavioral component of larval translational velocity vector
fluid component of larval translational velocity vector
observed larval translational velocity vector
larval terminal sinking velocity
kinetic energy dissipation rate
horizontal component of vorticity
Kolmogorov length scale
angle of larval propulsion relative to larval axis
von Karmann's constant (=0.41)
kinematic viscosity (=0.01 cm2 s−1)
Kolmogorov velocity scale
Kolmogorov time scale
particle response time
angle of larval axial rotation due to shear
No competing interests declared.