Introduction

Small copepods (≤1 mm) have been recognised to constitute a large portion of marine zooplankton, not only in terms of abundance but also biomass (Dam et al. 1993; Calbet et al. 2001; Hopcroft et al. 2001), and to play a relevant role in grazing (Morales et al. 1991; Turner 2004 and references therein). Small copepods provide the link between classic and microbial food webs (Stoecker and McDowell Capuzzo 1990; Kleppel 1993; Roff et al. 1995), and are prey for larval fish and other zooplankton consumers (Øresland 2000). Even when not substantially contributing to total biomass, as in the case of Arctic areas, small copepods can represent up to 25% of zooplankton production (Hopcroft et al. 2005). Their success is related to the variety of reproductive strategies and feeding habits sustaining their ubiquity and survival (Turner 2004 and references therein).

Clausocalanus furcatus (∼1 mm total length) occurs in tropical and subtropical areas of both hemispheres (Frost and Fleminger 1968) and is one of the most abundant calanoids in epipelagic waters of both oligotrophic (Webber and Roff 1995; Siokou-Frangou et al. 1997; Peralba and Mazzocchi 2004) and coastal eutrophic regions (Mazzocchi and Ribera d’Alcalà 1995). Although spatial and temporal distributions of this species are well documented, scanty information is available on its biology. A pioneering work by Mazzocchi and Paffenhöfer (1998) provided preliminary data on its feeding, reproduction and development. C. furcatus spends most of the time (73–100%) moving actively by a repetitive looping performed at high speed (10 mm s−1 on average), with frequent changes—and often complete reversal—of direction (Mazzocchi and Paffenhöfer 1999). These loops are occasionally interrupted by short but fast linear displacements or very fast somersaults, and sinking phases are rarely observed (Mazzocchi and Paffenhöfer 1999). This motion has also been observed in C. furcatus copepodids (our personal observations), while nauplii move forward by turning successive somersaults (Björnberg 1972). This swimming behaviour is displayed by other Clausocalanus species (Mazzocchi and Paffenhöfer 1999) but has not been reported for any other small copepod so far; also the nauplius movement differs from that of other calanoids (Björnberg 1972).

A sound comprehension of the ecological role of C. furcatus can be achieved only by merging and complementing the individual-level mechanisms of interaction (both with the environment and with other organisms) with the large-scale patterns observed in the field. Swimming behaviour can be utilised as an indicator of the adaptations of a species to the various environmental and internal conditions experienced. Swimming activity affects prey capture and is in turn modified by the alimentary conditions experienced (e.g. prey availability, food quality and/or quantity) (Williamson 1981; Tiselius 1992; Leising and Franks 2002). Feeding success will also depend on other factors, such as prey behaviour (Broglio et al. 2001) and escape ability (Yen and Fields 1992). The choice and selection of a specific foraging tactic (feeding currents, ambush predation) is inherently dependent on the swimming behaviour adopted, even though copepods can often apply various means of capturing prey to maximise the probability of successful captures (Landry 1980; Saiz and Kiørboe 1995; Kiørboe et al. 1996).

The present study was focused on the relationship between the swimming behaviour of C. furcatus and its feeding strategies, with the ultimate scope of understanding the success of this species in oligotrophic environments. It was aimed at: (1) characterising the swimming trajectories of C. furcatus in terms of their Lagrangian and fractal properties, (2) evaluating whether the occurrence of a capture resulted in any appreciable variation in the characteristics of copepods’ motion, (3) estimating the shape and extension of the predatory horizon. The analyses were based on video recordings of C. furcatus adult females in presence of dinoflagellate cells.

Materials and methods

Data set

The present work is based on the same data set partially utilised by Mazzocchi and Paffenhöfer (1999). C. furcatus adult females were sorted from zooplankton samples collected off the southeastern coast of Georgia, at the edge of the Gulf Stream. In the laboratory, the copepods were transferred to 1,920 ml glass jars and placed on a plankton wheel (0.2 r.p.m.). They were kept at 20°C, 14 light/10 dark hours, and food was represented by the dinoflagellate Gymnodinium nelsoni and the flagellate Rhodomonas sp., at a concentration of 0.2 mm3 l−1 for each species. These same conditions had allowed to rear up to four consecutive generations of this species (Mazzocchi and Paffenhöfer 1998).

A few days after the sampling, adult females were sorted from the rearing jars, checked under a dissecting microscope for their morphological integrity and health, and placed in 3 l vessels filled with copepod preconditioned filtered seawater (0.45 μm GF/F). The relatively large volume of the vessel was chosen to reduce the probability of encountering walls and a glass cover was placed to prevent turbulence. Eight free-swimming C. furcatus adult females were observed in darkness (utilising infrared light) with a modified CritterCam® (Strickler 1975) (60 fields s−1), with a field of view of 16 × 10 mm and a spatial resolution of ∼15 μm. Copepod swimming tracks were recorded in two-dimensions. Four different diets were offered during video observations (Table 1). Due to prey size and video equipment resolution, only dinoflagellates were clearly recognisable in the recordings; for this reason, all the considerations about the capture mechanisms are here referred to dinoflagellate prey. Copepods were left to acclimatize to the experimental conditions for 15–30 min before filming.

Table 1 Diets offered to the eight C. furcatus adult females recorded
Full size table

To investigate the Lagrangian and fractal features of C. furcatus swimming tracks, we selected all the clips in which a specimen was in focus for at least 2 s, allowing the analysis of the motion for a time span sufficiently long to calculate statistically significant measurements. The numerical analysis was applied to a total of 92 sequences, described by all eight specimens recorded at different times of the day. The sequences were divided into two classes: the first class (CPT: n = 36) included all the trajectories associated with the capture of a prey; the second class (SWM: n = 56) included sequences where the copepod swam close to a prey item, but without a capture (Table 2). The particular somersaulting behaviour that had been observed in C. furcatus by Mazzocchi and Paffenhöfer (1999) was not examined in the present work because somersaults have a strong three-dimensional component, therefore they cannot be properly analyzed using the two-dimensional video recordings available.

Table 2 Details of the clips, for both swimming and capture classes of motion, extracted and characterised in terms of Lagrangian and fractal properties
Full size table

The reconstruction of the predatory horizon of C. furcatus was based on the same video recordings, but sequences were selected with different criteria from those adopted for track analysis. Only video sequences where both the swimming copepod and one or more dinoflagellate cells were in focus for at least a few fields were digitised. A total of 111 sequences were selected and analysed, including 42 sequences where successful cell captures were observed.

Selected sequences were manually digitised using a software implemented by “e-voluzione” (s.r.l., Napoli), by which the position of rostrum of the copepod, taken as a reference point (Doall et al. 2002), was tracked for the entire time length of the clip. For the reconstruction of the predatory horizon, the positions of the tips of the first antennae of the copepod and the position of the prey were recorded as well.

Lagrangian analysis

As in Uttieri et al. (2004), we characterised the two-dimensional trajectories of C. furcatus in terms of three Lagrangian tools: velocity autocovariance, spectral analysis and reconstruction of kinetic energies.

Velocity autocovariance gives an estimate of the randomness of the motion by estimating for how long the velocity at a generic time step t influences later motion. We have calculated the velocity autocovariance for the two velocity components (u and v, respectively along X and Y) of the swimming trajectories of C. furcatus, velocities being computed using a central difference method. Along the X direction, the autocovariance (Kundu 1990) is given by:

$$ r(\tau ) = \frac{{\overline{{u(t)\,u(t + \tau )}} }} {{\overline{{u^{2} }} }} $$

where \( \overline{{u^{2} }} \) is the mean square velocity. The memory of the process is represented by the integral time scale T, given by (Kundu 1990):

$$ T = {\int\limits_0^\infty {r(\tau )d\tau } } $$

Low T values are indicative of processes resembling a random phenomenon. In the following, we will make use of the annotations T x and T y to distinguish the integral time scales associated respectively with the velocity components u and v.

Spectral analysis allows a description of the peak frequencies (if any) associated with the swimming motion. In this framework, the analysis of the time series is possible using the power spectral density function S(ω) (PSD henceforth):

$$ S(\omega ) = \frac{1} {{2\pi }}{\int\limits_{ - \infty }^\infty {{\text{e}}^{{ - i\,\omega \,t}} \,r(\tau )\,\overline{{u^{2} }} \,d\tau } } $$

describing the data in terms of their peak frequency composition (Emery and Thomson 1998). The PSD is the Fourier transform of the autocovariance r(τ); as a consequence, the results obtained by these two applications are complementary, giving a more exhaustive description of the analyzed phenomenon. In the present work, the PSD has been determined by applying the fast Fourier transform (FFT) (Press et al. 1989) to the velocity data series. The PSD function represents how the energy associated with the process is distributed over the spectrum of frequencies (Kundu 1990). The presence of peaks in the PSD corresponds to characteristic frequencies in the motion; conversely, the absence of a defined slope indicates that the motion derives from the superimposition of numerous frequencies, none of which being dominant over the others. In the latter case, the spectrum of a zero memory process is defined as “white” (Bendat and Piersol 1966).

The relative importance of the random component of the movement over the total kinetic budget can be assessed through the reconstruction of the kinetic energies. This was achieved by the evaluation, for each trajectory, of the ratio between the fluctuating part (EKE, eddy kinetic energy) and the total energy (TKE, total kinetic energy) of the motion. Based on Reynolds decomposition (Pond and Pickard 1983), the total velocity \( \underline{u} \) of a body can be considered as the sum of a mean field \( (\underline{{\overline{u} }} ) \) and of a superimposed stochastic one \( (\underline{{u'}} ): \)

$$ \underline{u} = \overline{{\underline{u} }} + \underline{{{u}\ifmmode{'}\else$'$\fi }} $$

where the underbar is indicative of vectorial notation. It is thus possible to define three distinct kinetic energies. The TKE per unit mass is:

$$ {\text{TKE}} = \frac{{\,\underline{{\,u}} ^{2} }} {2} $$

whereas the kinetic energies associated with the mean (MKE) and stochastic (eddy) (EKE) field are, respectively:

$$ {\text{MKE}} = \frac{{\,\,\,\underline{{\overline{u} }} ^{2} }} {2} $$
$$ {\text{EKE}} = \frac{{\,\,\,\underline{{{u}\ifmmode{'}\else$'$\fi }} ^{2} }} {2} $$

It is worth noticing that the above energies are kinematic properties associated with the organisms’ motion, and do not reflect any metabolic process. In this work, attention has been concentrated on the EKE/TKE ratio over the XY plane, by combining the two velocity components (Patterson 1985). The EKE/TKE ratio determines the relative importance of the random component of the velocity in the overall motion, and is therefore an estimate of the irregularity of the motion itself. Values close to unity are typical of random processes, for which most of the total kinetic budget is due to the stochastic part.

Fractal dimension

The degree of morphological complexity of a trajectory provides indirect information about the searching strategies of a copepod. A first parameter used to evaluate the complexity of a swimming path was the net-to-gross displacement ratio (NGDR) (Buskey 1984), representing the ratio of the linear distance between the beginning and the end of the path (net displacement) to the total distance (gross displacement) travelled by the organism. Despite its usefulness, NGDR has been demonstrated to suffer from scale-dependency (Dodson et al. 1997; Seuront et al. 2004a, b), and is therefore not always sufficiently precise. Fractal dimension (Mandelbrot 1967) overcomes this limitation: it provides precise and reliable estimates of morphological complexity without any scale-dependency (Coughlin et al. 1992; Seuront et al. 2004a, b; Uttieri et al. 2005).

In the present work, the fractal dimension D was estimated for C. furcatus swimming trajectories using a modification of the box-counting methodology proposed by Uttieri et al. (2005). This approach consists in superimposing to the track a square-elements grid and counting the number of cells occupied; this computation is repeated for different square lengths (r). The range of r values used in D estimation consists of a 2n sequence, and the lattice is placed with its lower and left margins tangential to the trajectory (as suggested by Buczkowski et al. 1998). All tracks were checked for the presence of an intermittent pattern in the displacements over different scales, indicative of a statistical self-similar fractal structure (Cole 1995), and a calibration of the technique was carried out with the aim of retrieving a correction factor necessary for the correct estimation of D values (Uttieri et al. 2005).

Throughout our analysis we did not focus on geometrical properties, such as curvature and torsion (Crenshaw et al. 2000) which, despite their usefulness in picking out trajectories specific features, do not provide further indications on the morphological complexity of a track.

Reconstruction of predatory horizon

Following Gerritsen and Strickler (1977), a capture of a dinoflagellate cell by C. furcatus was considered successful only when it was followed by the ingestion of the food particle. This was verified through a frame-by-frame analysis of the fields preceding and following the capture. Only when the dinoflagellate completely disappeared from the screen, the capture was considered successful.

Once the positions of the copepod and of the dinoflagellate were digitised, the coordinates were rotated to create a “Copepod-Centred Reference Frame” (henceforth CCRF), where the origin overlapped with the rostrum of C. furcatus, its first antennae corresponding to the x axis. The CCRF can be considered as an intrinsic reference frame (Beatty 1986), i.e. a frame which is not fixed in space but “follows” the item of interest. The CCRF allows visualising the position of the prey with respect to the head and to the first antennae of the copepod. The relative positions of captured and not captured prey with respect to the rostrum and first antennae permitted the reconstruction of the predatory horizon of C. furcatus.

An index of efficiency of capture (EC) was measured as:

$$ {\text{EC}} = \frac{{{\text{no}}{\text{. }}\,{\text{of}}\,\,{\text{captured}}\,\,{\text{prey}}}} {{{\text{no}}{\text{.}}\,\,{\text{of}}\,\,{\text{total}}\,\,{\text{items}}\,\,{\text{encountered}}}} \times \,100 $$

This parameter calculated the probability that an encounter could result in a successful capture, thus representing a proxy of the feeding performance of the copepod.

Statistical analyses

The results obtained were analysed using a one-way analysis of variance (ANOVA) (Sokal and Rohlf 1995) to evaluate if there was any difference in the two conditions analysed (CPT vs. SWM).

ANOVA was graphically complemented with the boxplot method (Tukey 1977; McGill et al. 1978). For each sample (i.e. one class of motion), a box and whisker plot is drawn: the bottom and the top of the box are respectively the 25th and 75th quartiles of the range, whereas the line inside it is the median. The whiskers extending from the two ends of the box show the extent of the rest of the data. Values beyond the end of the whiskers are represented with a cross; if no data are present beyond the box, a dot is drawn at the bottom of the whisker. The length of the box is the interquartile range, also referred to as dispersion (in a statistical sense). Two samples can be considered statistically similar when the median lines overlap each other and the dispersions are comparable.

The degree of fitting of the least-squares regression lines for fractal dimension evaluation was evaluated through the coefficient of determination r 2 (Sokal and Rohlf 1995).

Results

Swimming trajectories

The swimming trajectories described by Clausocalanus furcatus adult females were typically convoluted and rather elongated along the X direction (Fig. 1). In both classes of motion (SWM and CPT), the velocities were characterised by low autocorrelation, with small T values (SWM: T x  = 0.68 ± 0.32 s; T y  = 0.75 ± 0.38 s − CPT: T x  = 0.71 ± 0.27 s; T y  = 0.70 ± 0.25 s; mean value ± standard deviation). These results were statistically similar (ANOVA, P = 0.2918; boxplots in Fig. 2), indicating that the capture of a prey item did not affect the Lagrangian characteristics of the motion.

Fig. 1
figure 1

Examples of swimming trajectories described by freely swimming adult females of C. furcatus. The paths were obtained from digitised tracks, splined and subsequently smoothed for graphical reasons. Labels (s and f) indicate, respectively the starting and final points. All tracks shown were described in 8 s

Full size image
Fig. 2
figure 2

Boxplots of the integral time scale T evaluated for the two velocity components (T x for u, T y for v) of the trajectories described by C. furcatus in the two classes of motion considered in the present study, showing no statistical difference between the different T values (ANOVA, P = 0.2918) (SWM swimming, CPT capture)

Full size image

PSD plots (Fig. 3) show that the motion of C. furcatus can be considered as the superimposition of a large number of frequencies. No dominant peak was recognisable, with a white spectrum typical of processes lacking a relevant memory. As for the autocovariance, this result was maintained in both SWM and CPT, without any noticeable difference between the two classes of motion.

Fig. 3
figure 3

Example of power spectral density plot calculated on one velocity component of a C. furcatus swimming trajectory. No major peaks could be identified, indicating that the motion resembled a random process

Full size image

In accordance with the above-mentioned results, the determination of the EKE/TKE ratio revealed that, in both conditions, there was a relevant random component in the paths. The EKE/TKE values were in fact very close to unity (SWM: 0.97 ± 0.01 − CPT: 0.97 ± 0.02), indicating that more than 95% of the total kinetic budget of the motion was due to the random (eddy) component. Also in this case, results were statistically similar (ANOVA, P = 0.1878; boxplots in Fig. 4).

Fig. 4
figure 4

Boxplots of the EKE/TKE values for the two classes of C. furcatus motion considered, indicating no difference between them (ANOVA, P = 0.1878) (SWM swimming, CPT capture)

Full size image

The trajectories of C. furcatus presented an intermittent pattern in the displacements (Fig. 5), indicative of a fractal structure (Cole 1995). The swimming tracks of C. furcatus spread predominantly along one direction only, and the presence of a capture event did not determine any modification in the complexity of the trajectory. The tracks attained low fractal dimension values (SWM: D = 1.10 ± 0.08 − CPT: D = 1.14 ± 0.08), independently on the class of motion analysed (ANOVA, P = 0.2463; boxplot in Fig. 6). The reliability of the estimates was confirmed by the high r 2 values (SWM: r 2 = 0.96 ± 0.01 − CPT: r 2 = 0.97 ± 0.01).

Fig. 5
figure 5

Intermittent structure of the two-dimensional displacements of C. furcatus. The large (upper panel) and small (lower panel) scale patterns are statistically self-similar, indicating self-similarity over different spatial scales

Full size image
Fig. 6
figure 6

Boxplots of the D values for the swimming trajectories of C. furcatus associated (CPT) and not associated (SWM) with a cell capture event. The two classes of motion were not significantly different (ANOVA, P = 0.2463)

Full size image

Table 2 summarises the values, for each female digitised, of the different parameters evaluated in the present analyses.

Predatory horizon

Out of the 111 sequences digitised, only 42 resulted in a successful capture, with an overall efficiency of capture EC = 38.2%. In the remaining 69 clips, the copepod did not even attempt to grab a dinoflagellate. The video recordings showed that C. furcatus did not create any feeding current, nor did it display raptorial behaviour: all captures were performed by direct interception of the prey, when it was aligned along the anterior–posterior axis of symmetry of the copepod (Fig. 7), in proximity of its head. On capturing a dinoflagellate cell, C. furcatus did not change its motion, but kept on swimming according to its peculiar behaviour, with the first antennae wide stretched. Only on a few occasions, prey capture was observed from the lateral view: in these cases, handling of the cells by copepod’s mouthparts was observed, sometimes followed by cell rejection rather than ingestion.

Fig. 7
figure 7

Two-dimensional pattern of predatory horizon in the copepod-centred reference frame for C. furcatus (prey and copepod not in scale). All captures were concentrated in a rather restricted area frontal to adult females. For captured cells, the position immediately preceding the capture (one frame, 1/60 s) before grabbing is reported, whereas for not-captured cells the position closest to copepod’s rostrum is considered. It is worth noticing that in the plane of reference the position of some prey overlapped, therefore the number of items reported in the schematic is less than the number of interactions investigated (111). Labels identify areas of the predatory horizon with different efficiency of capture (see text for details)

Full size image

The predatory horizon of C. furcatus has been estimated as an area placed frontally to the copepod (Fig. 7), extending 0.4 mm frontally and 0.2 mm laterally to the head of the copepod, with an EC = 84%. This area can be divided in two parts: a central corridor (A), approximately 0.06 mm2 wide, with a EC = 100%; and an external area (B) with EC = 62%. Outside this region, no capture took place (EC = 0%). No multiple catches were observed, even when two or more cells were very close together and all inside the predatory horizon.

Discussion and conclusions

Perceiving and capturing a prey results from a proper localisation and identification of the item, and from a successful predatory habit. Whilst for the first process a major role is played by the mechano- and chemo-sensory abilities of the copepod, for the second one the swimming behaviour has noticeable influence on capture dynamics. In this framework, the analysis of the video recordings of freely swimming copepods in presence of food items provides important details on the mechanisms of interaction between a copepod and its prey, at the scales proper to the small organisms involved.

Swimming trajectories and feeding strategy

The numerical characterisation of the swimming paths performed by C. furcatus provides useful insights into the properties associated with the small-scale motion of this species, and consequently on its behaviour. The peculiar motion of C. furcatus is maintained over the entire day without any significant individual variability, as instead noted in other species (Turner et al. 1993; Seuront et al. 2004b). The Lagrangian description shows that the motion of C. furcatus has properties recalling a random walk, i.e. a relevant random component and absence of dominant frequencies in power spectral density plots. These features did not change during the capture of a prey, and can therefore be considered as the typical and persistent characteristic of the swimming behaviour of this species. The Lagrangian techniques here utilised have been first applied to the three-dimensional trajectories described by the freshwater cladoceran Daphnia pulex (Uttieri et al. 2004). By comparing the results obtained for C. furcatus and D. pulex, some interesting common features emerge. Despite the two organisms show qualitatively different swimming behaviour, their kinematic properties are strikingly similar. In both cases, the motion has random-like properties, suggesting that these kinematic properties might be considered to be a common strategy adopted by zooplankters to live and perform in the water column, independent of the morphology of their tracks. The random properties and the lack in a peak frequency in the PSD plots can be interpreted as a predator-avoidance technique by means of a “fluid-mechanical camouflage” (Hwang and Strickler 2001), as already put forward also for D. pulex (Uttieri et al. 2004). These properties may in fact allow the copepod to confuse potential predators, which may thus lack clearly distinguishable cues to localise and capture it. This would explain the low predatory rates by chaetognaths on C. furcatus reported by Kehayias et al. (1996), as chaetognaths’ feeding bouts are triggered by specific frequencies (Feigenbaum and Reeve 1977) which are not displayed by C. furcatus. Kehayias et al. (1996)’s results are further confirmed by unpublished data by P. Simonelli and M. G. Mazzocchi, who carried out a study on the gut contents of Sagitta collected in the Gulf of Naples over a 3 years period. Their results show that, even during the peak of abundance of C. furcatus, this copepod is preyed at negligible extent by Sagitta, differently from the co-occurring calanoid Temora stylifera, which displays a different cruising swimming behaviour. Consequently, the mechanical self-hiding mechanism proposed above for C. furcatus would counteract the strong signals likely created by the copepod itself owing to its continuous motion (as discussed in Paffenhöfer and Knowles 1980, for copepod nauplii).

The low values of fractal dimension indicate that the tracks of C. furcatus have a reduced tendency to occupy the available space, but indeed remain confined to a restricted area. This represents a confirmation of the small-volume exploitation strategy proposed by Mazzocchi and Paffenhöfer (1999), recalling the area-restricted search foraging strategy (Tinbergen et al. 1967) adopted by different organisms (as reviewed in Leising 2002). As discussed in Uttieri et al. (2005), it is not possible to get three-dimensional information basing on two-dimensional data; however, modelling results by Wiggert et al. (2005) show that also in a three-dimensional reference frame the tracks of C. furcatus consist of a series of continuous loops alternated with straight vertical movements, confirming the sequential investigation of small volumes even in a three-dimensional reference frame, exploring neighbouring volumes one after another.

Clausocalanus furcatus captures only those dinoflagellates placed within a well-defined predatory horizon, outside of which no capture takes place independently of the distance from its first antennae. Such a feeding strategy presents marked differences with the capture habits displayed by other species. While ambush predators maximise their sensory field by attacking prey perceived by means of their mechanoreceptors (e.g. Oithona similis, Svensen and Kiørboe, 2000; O. plumifera, Paffenhöfer and Mazzocchi 2002), C. furcatus possesses a small predatory horizon, especially if compared to that of other actively swimming copepods (e.g. Cyclops: Kerfoot 1978; Euchaeta rimana: Doall et al. 2002). Additionally, C. furcatus does not create feeding currents, differently from other small calanoids displaying cruising behaviour, such as Paracalanus aculeatus (Paffenhöfer 1998), nor does it perform feeding bouts to catch distantly located prey.

The emerging scenario is quite different from the classic four-step interaction (detection–recognition–capture–ingestion) discussed in Gerritsen and Strickler (1977). In the case of C. furcatus, the copepod appears to sense only those prey in the close vicinity of its rostrum apparently collecting them almost like a passive sieve. However, the observed cell handling and occasional rejection seem to indicate a selection likely by the second antennae and mouthparts due to palatable versus unpalatable prey.

The observed captures of cells extend to about 200 μm to the left and right of the rostrum. That is within the range of the setae of the second antennae (A2) which, according to measurements of C. furcatus female appendages, extend with their tips to more than 0.45 mm from the median of the cephalon. As the A2 move in an approximate circular manner, as observed e.g. for Eucalanus pileatus (e.g. Paffenhöfer and Lewis 1990), they cover with the tips of their setae a large volume of water per unit time, probably operating not slower than 40 Hz (unpublished data for Temora turbinata). Thus, the numerous apical-pore setae that C. furcatus females bear on their A2 (our personal observations), as already reported by Paffenhöfer and Loyd (2000) for other species (Centropages velificatus, E. pileatus, P. aculeatus and Temora stylifera), could perceive signal-producing chemical compounds of phytoplankton cells as they (A2 and also other moving cephalic appendages) move continuously. While the reaction times to hydrodynamic signals are near 2.5 ms (Lenz and Hartline 1999), those to chemical signals appear longer (Jiang et al. 2002). From our unpublished observations we can approximate that the reaction time to chemical compounds is not longer than 10 ms, in agreement with the numerical results from Jiang et al. (2002). Within 10 ms a C. furcatus female has travelled not more than 100 μm and thus should be able to capture the perceived cell with its cephalic appendages.

From the abovementioned data one could calculate the approximate volume of water which would be scanned by the copepod’s A2 while moving forward. As the width of the path in which cells are perceived is about 0.4 mm, and the distance to which the tips of the A2 extend ventrally is at least 0.45 mm, the area which is covered ventrally while moving forward is 0.4 × 0.45 mm = 0.18 mm2; multiplied by the velocity of 10 mm s−1 the volume of water scanned is 1.8 mm3 s−1. This would result in 156 ml scanned 24 h−1. This volume is in the range of the average maximum clearance rate of females of C. furcatus feeding on in situ particulate matter at 23°C, which was 218 ml female−1 day−1 (Paffenhöfer et al. 2006). Furthermore, C. furcatus’ oxygen consumption measurements carried out by Paffenhöfer (2006) show that these clearance rates at oceanic particle concentrations are sufficient to cover the metabolic needs of this continuously moving copepod.

Our video recordings have been performed in still water conditions, consequently all conclusions can only been referred to similar situations and not in presence of turbulence. It is well known however that small-scale turbulence may affect the behaviour and the encounter rates of copepods with potential prey items (Rothschild and Osborn 1988). It is reasonable to guess that, in presence of such an environmental signal, C. furcatus’ behaviour may be altered, at least at the onset of turbulence, until some degree of habituation is reached (as reported for other species in Costello et al. 1990) allowing the copepod to recover its original motion. Future experiments will test this hypothesis.

Interacting with the environment

The shape and limited extension of the predatory horizon, as well as the direct interception of food particles, depend on the motion behaviour of C. furcatus. The frantic and very fast motion does not allow the copepod to create feeding currents, for which a smoother natatory activity is required (e.g. Bundy et al. 1993). In addition, due to its fast propulsion the copepod does not seem to be able to veer rapidly towards prey located outside the predatory horizon. Such an attack would consist of four consecutive steps (identification and localisation of the quarry, reorientation towards the prey and bout) occurring over very short time scales (50–100 ms) (Bundy et al. 1998; Doall et al. 2002; Paffenhöfer and Mazzocchi 2002).

It appears that the swimming activity and the feeding strategy of C. furcatus are closely related. How can such behaviour allow the copepod to thrive in oligotrophic regions? To answer this question, it is necessary to investigate which are the environmental conditions supporting the peaks of this species.

Clausocalanus furcatus is a dominant species in tropical and subtropical oligotrophic regions of both hemispheres (Frost and Fleminger 1968). It is present in the upper part of the water column and the core of the population is always above the thermocline (Fragopoulu et al. 2001; Paffenhöfer and Mazzocchi 2003; Peralba and Mazzocchi 2004). Long-term studies carried out in the Mediterranean show that C. furcatus has its main peak of abundance in late summer–autumn (Mazzocchi and Ribera d’Alcalà 1995), when the physical structure of the water column is favourable to the development of phytoplankton micro-layers (e.g. Dekshenieks et al. 2001; McManus et al. 2003, 2005). As underlined in Seuront et al. (2001), predator–prey encounter rates are crucially dependent on behavioural adaptations to prey patchiness; consequently, micro-layers may have a potentially significant ecological role. Recent studies show that copepods can detect aggregates chemically rather than hydromechanically (Kiørboe 2001; Kiørboe and Thygesen 2001). Such patches can also be efficiently utilised by protists (e.g. Oxyrrhis marina; Menden-Deuer and Grünbaum 2006), with positive effects on their feeding efficiency.

The continuous looping swimming can be considered an optimization for the detection of micro-scale patches of food (Uchima and Hirano 1988). C. furcatus moves along convoluted but restricted tracks, which may be considered as an optimisation to a inhomogeneous landscape. A dense cluster of particles might thus be advantageous for C. furcatus, which could remain inside the patch and thoroughly exploit it enhancing the probability of encountering new prey items.

Small-scale processes contribute to large-scale dynamics in the oceans (Dodson et al. 1997). Investigating the interactions occurring at the individual scale is therefore fundamental for understanding the role played by copepods in the functioning of pelagic systems. Our results have shown that the atypical swimming behaviour of the calanoid C. furcatus is coupled with restricted predatory horizon and direct interception of prey, likely representing an efficient adaptation to thrive in oligotrophic areas of all seas.