[13] released virtual particles at two locations in the surface layer of Global Hybrid Coordinate Ocean Model (HYCOM) output [28] using the particle-tracking software Ichthyop v. 3.2 [29]. Interestingly, the distribution of turtles in figure1a of Briscoe et al. We then assessed whether particle speeds were correlated with the speed of the drifter using Spearman's rank-order correlation and whether particle directions were correlated with the direction of the drifter using CircularCircular correlations. Raising sea turtles for up to 3 years prior to release and developing satellite telemetry attachment methods that lasted between 173 days and 865 days are remarkable achievements. Tracking virtual particles offshore of Japan shows the potential for substantial variation in transport (figure1a). Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. We used Ichthyop v. 2 software [29] to release 1 000 virtual particles at the same deployment locations and dates as the 44 turtles in Supplemental table 1 of Briscoe et al. Whatever analytical approach is adopted, it should be paired with an identical analysis applied to the tracks of passive oceanographic drifters (figure3) [4]. A similar analysis was performed in which 500 virtual particles were released in a 0.08 0.08 rectangle centred on the drifter's initial position and tracked for 230 days.
A daily location was recorded for each particle and the number of particles within each 1 1 grid cell was summed across each day of the simulation. Figure1 of their paper implies the particle release was substantially eastward of the 9 April 2010 turtle release and southward of the 12 July 2011 turtle release(s). The depths that turtles frequent likely vary among oceanic areas [34] and the impacts from winds (such as storms) may be more pronounced in certain areas and times [35]. The spatial heterogeneity in maximum eastward transport (figure1a) highlights the importance of co-localizing measures of ocean currents with tracking data. those corresponding to specific dates) compared with those summed over the 36 years of drifter data. Given the quality of the datasets that they are working from, robust analyses could be achieved with simple modifications to their present methods. [13] compared the track velocities of turtles with the velocities of virtual particles released at a site distant to that of the turtle release sites. Marine animal behaviour: neglecting ocean currents can lead us up the wrong track, A biologist's guide to assessing ocean currents: a review, Animal orientation strategies for movement in flows, The magnetic map of hatchling loggerhead sea turtles, Simulating transoceanic migrations of young loggerhead sea turtles: merging magnetic navigation behavior with an ocean circulation model, Orientation behavior in fish larvae: a missing piece to Hjort's critical period hypothesis, A little movement orientated to the geomagnetic field makes a big difference in strong flows, Ontogeny in marine tagging and tracking science: technologies and data gaps, Current-oriented swimming by jellyfish and its role in bloom maintenance, Direct evidence of swimming behavior demonstrates active dispersal in the sea turtle lost years., Active dispersal in loggerhead sea turtles (, New perspectives on the pelagic stage of sea turtle development, Active swimmerspassive drifters: the oceanic juvenile stage of loggerheads in the Atlantic system. Early on, ocean currents were presumed to dominate organisms movements, owing to limited swimming capacity relative to ocean velocity and/or a limited ability of animals to direct their swimming in the barren sensorial-environment of the open sea [2]. Colours in each plot are log10-scaled relative to the HYCOM simulations, assuming a release of 1 000 virtual particles per release site, summed daily throughout the simulation. This suggests that the discordance in large-scale distribution between modelled currents and turtle tracks reported by Briscoe et al. The reported locations of release differ between their methods section and electronic supplementary material. [13], respectively. the turtle tracked for 865 days corresponded to an 870 days drifter simulation). Track durations ranged from 173 days to 865 days (mean = approx. [4345]). [13] would travel further east, even if turtles were entirely passive (figure1a,b). Only 19% of particles were predicted to cross into the Western Hemisphere after 469 days (figure1b), whereas 64% of particles did when allowed to drift for 865 days (figure1a). 
In all panels, circles indicate the release sites of turtles, stars are the approximate position of particle release, inferred from the supplementary information and from figure 1 of Briscoe et al. The authors could then assess a number of useful metrics including (i) ocean velocity along track segments, (ii) swimming velocity (subtracting ocean velocity from track velocity) along track segments, and (iii) separation distances between particles and the track through time. 
Briscoe et al. Furthermore, physical processes not characterized in ocean circulation models can result in substantial departures between predictions and the actual movements of a passive object in the ocean (figure3). Distribution of genetic diversity reveals colonization patterns and philopatry of the loggerhead sea turtles across geographic scales, The establishment of a pelagic Sargassum population in the tropical Atlantic: Biological consequences of a basin-scale long distance dispersal event, Observation and quantification of inertial effects on the drift of floating objects at the ocean surface, Nearshore neonate dispersal of Atlantic leatherback turtles (Dermochelys coriacea) from a non-recovering subpopulation, Improving transport predictions of pelagic Sargassum, Assessing reliance on vector navigation in the long-distance oceanic migrations of green sea turtles. Therefore, before conclusions can be reached about the role of swimming behaviour on the movements of animals, it must first be shown that differences between virtual particles and the animal's track are greater than the differences between virtual particles and the tracks of drifters within the same region. Small dots indicate the final locations of virtual particles. We tested the ability of the approach to characterize the velocity of a passively drifting object by releasing 500 virtual particles in a 0.08 0.08 rectangle that was 0.5 in latitude to the south of the drifter's start position. The analyses performed on individual tracks could then be aggregated to gain population-level insight into the movements of turtles [12,40]. Can drifting objects drive the movements of a vulnerable pelagic shark? By contrast, releasing particles at the start location of the drifter (figure3c,d) or some distance away (figure3a,b) resulted in ocean velocity estimates unrelated to drifter speed (Spearman's r = 0.002, p = 0.978, n = 230; Spearman's r = 0.096, p = 0.149, n = 230, respectively) or drifter direction (CircularCircular correlation r = 0.06, p = 0.365, n = 230, CircularCircular correlation r = 0.023, p = 0.729, n = 230, respectively). In each of the above cases, particle trajectories were computed at 30 min intervals using the RungeKutta fourth-order time-stepping method and recorded daily.
Drifters are deployed with drogues (i.e. [13] are the result of differences between particle release sites and sites of turtle deployments. Anguilla rostrata [13] obtained data on turtle movement by laboratory-rearing 44 loggerhead sea turtles to an age of 13 years (29.737.5 cm straight carapace length), outfitting the turtles with satellite transmitters and releasing them on two separate days, 9 April 2010 (n = 17) and 12 July 2011 (n = 27). In previous papers, we have argued that natural selection should favour those organisms that bias locomotion in directions that, on average, lead to favourable areas (e.g.
In this way, a wide variety of ocean conditions are accounted for, but only those most closely matching the turtles' movements are used in statistical analyses [40]. To determine what influence these different start dates might have on transport predictions, we performed the same analyses described in 2b and c, but for the day before and after the turtle release date (days 8 and 10 April 2010, and 11 and 13 July 2011). If the authors wanted to consider uncertainty in time (as in their initial analyses) or depth, they could release particles from those locations at some time before and after the recorded occurrence and across a range of depths. The ocean model output used in this paper (1/12 global HYCOM + NCODA Ocean Reanalysis) was funded by the U.S. Navy and the Modeling and Simulation Coordination Office. From each start location, the maximum eastward longitude was determined for 865 days (the maximum turtle track duration and the duration of particle transport chosen by Briscoe et al. More recently, Briscoe et al. Briscoe et al. We declare we have no competing interests. (c) Same as in (a), but showing tracks of 500 virtual particles released at the start location of the drifter (pale blue).
To the east of 137.5 E, 10 000 virtual particles were released from random locations on 12 July 2011. The output is publicly available at http://hycom.org. Panels (a,b) represent predictions taking into account 36 years of in situ oceanic conditions, whereas panel (c) depicts modelled conditions that occurred during the tracking experiment by Briscoe et al. [13]. First, the central aim should be to adequately quantify ocean velocity over the area in which the turtle occurs. Predictions of transport vary considerably over distances of a few km and within 24 h periods. Less dispersion in the HYCOM predictions can be primarily attributed to fewer potential movement pathways realized under the unique release conditions in HYCOM (i.e. Briscoe et al. Here, we use a combination of modelling and in situ data to demonstrate that the magnitude that each of the above parameters varied within the study by Briscoe et al. Regardless, our results indicate that swimming behaviour should not immediately be assumed as the explanation for systematic differences between the movements and distributions of marine organisms and predictions based on ocean circulation models (figure3) or, necessarily, estimates derived from in situ measurements of ocean currents (figure2a,b). Here, we show that the magnitude of variation in physical parameters between turtles and virtual particles can profoundly alter transport predictions, potentially sufficient to explain the reported differences without evoking swimming behaviour. However, the experimental design implicitly assumed that transport predictions were insensitive to (i) start location, (ii) tracking duration, (iii) depth, and (iv) physical processes not depicted in the model. Passive drift or active swimming in marine organisms? Likewise, a significant correlation was found between daily drifter direction and the directions of particles released along the track (CircularCircular correlation r = 0.299, p < 0.001, n = 230). [13]) and tracked for 230 days. 
If the address matches an existing account you will receive an email with instructions to reset your password. (A second simulation was also performed, in which particle trajectories were computed for 4 years, to test whether transpacific transport was possible and was not used in direct comparison with the tracking data for the turtles.) Tracking data were examined from 18 June 2011 (37.044 N, 132.458 W) through 2 February 2012 (37.276 N, 152.730 W) at daily intervals (00.00 h GMT).
every 2, 5, or 10 days). [13], we performed analyses using the surface layer of Global HYCOM. These analyses indicate that care should be exercised when choosing an analytical approach to infer behaviour from tracking data, as approaches for quantifying ocean currents are not equivalent. However, turtle release locations were all seaward of the continental shelf (200 m isobath; figure1) and thus ocean currents at those sites should be depicted reasonably well by Global HYCOM. Drifting with Flow versus Self-MigratingHow Do Young Anadromous Fish Move to the Sea? We encourage future research in this area to employ robust experimental design that uses multiple methodsmodelling in the context of in situ observations of ocean circulation and organismal movement [11,12]. Particle velocity estimates could be obtained at daily intervals or, to reduce spatio-temporal autocorrelation (and artificial inflation of sample size for subsequent statistical analyses), some subset of the original tracking data (e.g. Panels (a,b) represent predictions taking into account 36 years of in situ oceanic conditions, whereas panel (c) depicts modelled conditions that occurred during the tracking experiment by Briscoe et al. In the case of identifying the role of ocean currents on the transport of a specific turtle, ocean conditions that are typical or representative are not particularly useful as they will tend to average-out the unique oceanic conditions (local weather, tidal phase, etc.) Predictions of maximum eastward transport for particles released from the same locations but a day apart indicate that an increase or decrease of eastward movement by 30 longitude is possible over a 469 or 865 day tracking period (figure1cf). For simplicity of presentation and to make results most relevant for comparison with Briscoe et al. To test the influence of track duration on transport predictions, we performed the same analyses described in 2b, but computed the maximum eastward longitude of each particle's trajectory after 469 days, the mean turtle track duration in Briscoe et al. [13], it appears that if turtles remained at the ocean surface for extended periods, eastward transport would be enhanced, but spending a greater portion of time at depth would increase retention in the western Pacific.
