Finding the ‘lost years’ in green turtles: insights from ocean circulation models and genetic analysis rspb.royalsocietypublishing.org
Nathan F. Putman1 and Eugenia Naro-Maciel2 1
Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, OR 97330, USA Department of Biology, College of Staten Island, City University of New York, 2800 Victory Boulevard, Staten Island, NY 10314, USA 2
Research Cite this article: Putman NF, Naro-Maciel E. 2013 Finding the ‘lost years’ in green turtles: insights from ocean circulation models and genetic analysis. Proc R Soc B 280: 20131468. http://dx.doi.org/10.1098/rspb.2013.1468
Received: 7 June 2013 Accepted: 15 July 2013
Subject Areas: ecology, environmental science, genetics Keywords: dispersal, distribution, marine turtles, Chelonia mydas, ocean currents, population structure
Author for correspondence: Nathan F. Putman e-mail: [email protected]
Electronic supplementary material is available at http://dx.doi.org/10.1098/rspb.2013.1468 or via http://rspb.royalsocietypublishing.org.
Organismal movement is an essential component of ecological processes and connectivity among ecosystems. However, estimating connectivity and identifying corridors of movement are challenging in oceanic organisms such as young turtles that disperse into the open sea and remain largely unobserved during a period known as ‘the lost years’. Using predictions of transport within an ocean circulation model and data from published genetic analysis, we present to our knowledge, the first basin-scale hypothesis of distribution and connectivity among major rookeries and foraging grounds (FGs) of green turtles (Chelonia mydas) during their ‘lost years’. Simulations indicate that transatlantic dispersal is likely to be common and that recurrent connectivity between the southwestern Indian Ocean and the South Atlantic is possible. The predicted distribution of pelagic juvenile turtles suggests that many ‘lost years hotspots’ are presently unstudied and located outside protected areas. These models, therefore, provide new information on possible dispersal pathways that link nesting beaches with FGs. These pathways may be of exceptional conservation concern owing to their importance for sea turtles during a critical developmental period.
1. Introduction Perhaps one of the greatest obstacles in providing a scientific basis for conservation and management of marine species is the lack of basic information on their distribution across life-history stages [1]. For instance, many marine taxa have remarkable fidelity to specific reproductive areas, yet disperse widely over vast expanses of the ocean to forage [2]. Thus, processes that occur in one location may drive changes at distant sites or in other ecosystems through the ecological coupling of long-distance movement [3–5]. Gathering such information has proved difficult for species that spend much or all of their time in the open sea, where direct observation is hampered by logistical constraints [4–8]. This is particularly true of young green turtles (Chelonia mydas), which rapidly disperse from natal beaches into the ocean. Observations of these turtles are largely lacking during the subsequent life stage termed ‘the lost years’, and they are generally only seen again three or more years later upon recruiting to near shore foraging grounds (FGs) as juveniles [7,8]. Here, we used a modelling approach to predict the oceanic distribution of juvenile green turtles during their ‘lost years’ throughout the Atlantic Ocean, Mediterranean Sea and the southwestern Indian Ocean. We used complementary techniques in which thousands of virtual particles were released into an ocean circulation model in the vicinity of major rookeries and known FGs (figure 1a). Particles released near rookeries were tracked forwards through time to estimate likely dispersal routes, whereas particles released near FGs were tracked backwards through time to identify possible oceanic corridors available to turtles for reaching their FGs. By tracking passively drifting particles from their sources (rookeries) and endpoints (FGs), we bracket the range of likely dispersal pathways and diminish potential errors associated with our lack of information on post-hatchling swimming behaviour [9].
& 2013 The Author(s) Published by the Royal Society. All rights reserved.
30° N 15° N 0° 15° S
25 000 1800 200
30° S
2. Material and methods
30° N 15° N 3.6
15° S 30° S
density
0°
2.8 2.0 1.2 0.4
45° S (c) 45° N 30° N 15° N
5 4
15° S
3
30° S 45° S
age
0°
2 1
90° W
60° W
30° W
0°
30° E
60° E
Figure 1. Maps of model domain and predicted green turtle distribution. (a) Locations of FGs (green circles) and rookeries (white squares) examined in this study. Sizes of squares are scaled to relative rookery size; abundance at FGs is unknown. (b) Predicted distribution of particles released at rookeries in proportion to population size. Colours indicate particle density within a grid cell throughout the simulations (counted daily, logarithmic scale). Orange-red areas indicate possible ‘lost years hotspots’ where turtles might be found in high densities. (c) Mean age of particles (in years) within each grid cell (counted at daily intervals) throughout the 5-year simulations.
We compared our modelling results to estimates from published genetic mixed-stock analyses (MSAs) [10–12]. The ‘many-to-many’ (m2m) MSA method, an extension of the original ‘one-to-many’ (o2m) MSA, uniquely incorporates metapopulation structure of multiple FGs and rookeries through two approaches that correspond to the two particle tracking techniques described earlier [11–13]. In FG-centric MSAs, the proportion of turtles at a given FG, which comes from each source population (rookery), is estimated based on mitochondrial control region sequences. In rookery-centric MSAs, the proportion of turtles from a given rookery, which go to different FGs, is estimated. Confidence intervals (CIs) for MSA estimates are typically wide owing to inherent limitations: some locations are unavailable for sampling and
We tracked the dispersal of virtual particles throughout the Atlantic and southwestern Indian Oceans, and the Mediterranean Sea, within hindcast output from the Global Hybrid Coordinate Ocean Model (HYCOM) (http://hycom.org). HYCOM output has a spatial resolution of 0.088 (approx. 6 –9 km grid spacing), a daily time step, and assimilates satellite and in situ measurements to produce realistic velocity fields [18]. Numerical experiments were performed with ICHTHYOP (v. 2) particle tracking software [19]. For advection of particles, ICHTHYOP implemented a Runge–Kutta fourth-order time-stepping method whereby particle position was calculated each half-hour. To simulate sub-grid scale turbulent processes, horizontal dispersion was also included in the advection process [19]. The HYCOM-ICHTHYOP system accurately predicts the movement of surface-drifting buoys [20] and is well established for studying sea turtle dispersal [6,9,20–23]. Young turtles are relatively weak swimmers and divers, and although even minimal swimming can influence the distribution of marine organisms [9,24], these turtles’ net movement is largely driven by ocean currents [6,12,15]. Additionally, simulating swimming behaviour would introduce uncertainty that, without a priori information on orientation behaviour of different populations (e.g. [23]), would be impossible to parametrize. Likewise, we did not specify ‘recruitment’ (halting movement after a certain age or location encountered, e.g. [21]) to allow a full range of transport possibilities to be mapped. Dispersal was simulated from 28 rookeries (2.582.58 zones) in the Atlantic and southwestern Indian Ocean, and the Mediterranean Sea (see figure 1a and the electronic supplementary material, table S1). Fifty particles were released daily during the peak 90 days of each rookery’s hatching season in 2004, 2005 and 2006 (a total of 13 500 particles per rookery; electronic supplementary material, table S1). Each particle was tracked forwards through time for a total of 5 years. We recorded the percentage of particles that entered 32 FGs (58 58 zones) that are known to host green turtles (figure 1a). To estimate the distribution of pelagic juvenile green turtles, we plotted the number of particles within each grid cell throughout the 5-year simulations and the mean age of particles within each cell. The number of particles plotted was weighted by an estimate of population size (number of nesting turtles) at the source rookeries (see the electronic supplementary material, table S1). To further explore connectivity between rookeries and FGs, 100 particles were released every 5 days at the 32 FGs from the previous analysis. Releases occurred from 1 January 2012 through to 1 January 2009 (i.e. over 3 years for a total of 21 900 particles per FG). Particles were tracked backwards through time for a total of 5 years. Thus, ICHTHYOP determined where a particle came from to reach its final location in the FG. We recorded the percentage of particles that entered the 28 rookeries defined in the previous analysis. We weighted the particle tracking results based on rookery population size to determine the contribution of each rookery to a given FG. We determined whether results from particle tracking simulations were correlated with m2m MSA estimates from Naro-Maciel et al. [10] (i.e. 13 rookeries and 13 juvenile FGs). Because the source location of all particles is known in our
Proc R Soc B 280: 20131468
45° S (b) 45° N
2
rspb.royalsocietypublishing.org
there are often broadly shared haplotypes among rookeries [6,13]. Thus, considering population connectivity based on physical transport processes is probably necessary to help guide MSA interpretation [12,14–17]. Here, the combination of oceanic dispersal simulations and genetic MSA estimates offer a substantive hypothesis regarding basin-wide population connectivity and provide unprecedented information on the distribution of young turtles during their ‘lost years’.
(a) 45° N
(a) Rookery-centric estimates of connectivity Plotting the predicted distribution of virtual particles in proportion to rookery size indicated spatial variation in particle abundance and age structure across the Atlantic (figure 1). In the North Atlantic, a high abundance of particles was predicted from the southwestern Caribbean Sea through the Gulf of Mexico along the eastern United States coastline to North Carolina (approx. 378 N) and throughout the Sargasso Sea (approx. 20–358 N and 75–308 W). A more localized abundance of particles was predicted offshore of Guinea Bissau (near the west coast of Africa, approx. 128 N). In the Mediterranean, particle hotspots were restricted to the eastern basin. In the South Atlantic, high density was predicted between approximately 158 S–308 S and west of 308 W; to the southwest of this region a less dense cluster of particles was also predicted at approximately 308 S and 158 W. In the southwestern Indian Ocean, a substantial number of particles was predicted throughout the Mozambique Channel and just south of 308 S between 308 E and 758 E. The percentage of particles forward-tracked from rookeries to FGs was significantly correlated with the rookery-centric mean MSA estimates for five of the 13 rookeries (r . 0.74, p , 0.003) (see the electronic supplementary material, table S2).
(b) Foraging ground-centric estimates of connectivity The per cent contribution of rookeries to a particular FG was estimated by both backtracking particles from FGs and forward-tracking particles from rookeries (see the electronic supplementary material, tables S3 and S4). Simulations indicated that a high degree of connectivity is possible among rookeries and FGs. For example, backtracking simulations indicated that particles from 17 rookeries passed through the Barbados FG, with approximately 26% arriving from Suriname and approximately 22% from Guinea Bissau (rookeries on opposite sides of the Atlantic; electronic supplementary material, table S2). Where FGs are in close proximity to a large rookery, however, contributions may be dominated by a single source. For instance, the Nicaragua FG was predicted to have 85% of particles arrive from the Costa Rica rookery based on backtracking and 94% of particles estimated by forward-tracking simulations. The weighted per cent contributions of particles that were backtracked from FGs to rookeries were well correlated with FG-centric MSA estimates for nine of the 13 FGs (r . 0.59, p 0.032) (see the electronic supplementary material, table S5). The weighted per cent contributions of particles that were forward-tracked from rookeries to FGs were well correlated with FG-centric MSA estimates for 10 of 13 FGs (r . 0.57, p 0.033) (see the electronic supplementary material, table S6). Both particle tracking techniques were
4. Discussion Our findings provide critical predictive information on the ‘lost years’ of green turtles. Determining where marine turtles spend their first years of life has been a fundamental problem of sea turtle ecology for decades [6–8]. Our findings represent, to our knowledge, the first quantitative and testable hypothesis regarding the oceanic distribution and age structure of young green turtles throughout an entire ocean basin and beyond. Underscoring the biological and conservation significance of the results, this methodology can be used to inform demographic models that require spatially explicit and age-based information, and can be readily applied to other marine species. Most of the area covered by particles and many of the oceanic regions identified as possible ‘lost years hotspots’ for green turtles do not correspond to previously described FGs (figure 1b) and may be difficult to manage because they are outside of the Exclusive Economic Zone of sovereign nations. Further complicating management, simulations indicate that very different dispersal trajectories are likely for turtles departing from the same rookery (see figure 2 and the electronic supplementary material, table S1). For instance, particles dispersing from Ascension Island were transported to the North Atlantic, South Atlantic, and Indian Oceans (figure 2b). Likewise, backtracking simulations suggest that 26 of the 32 FGs could have inputs from rookeries across these three basins. Genetic surveys of southwestern Indian Ocean rookeries revealed patterns consistent with limited migration between the Atlantic and Indian basins [25,26], lending support to our findings. Mixing of diverse genetic stocks has also been documented in FGs where older turtles occur [10,13,27]. Thus, evidence increasingly suggests that the management of many turtle populations will require a broad approach across life-stages and regions [6,13]. Dispersal patterns (figures 1, 2 and 3) were consistent with previous genetic and oceanographic studies that suggest considerable connectivity is possible among green turtle populations during the ‘lost years’ [12,15,25–27]. Dispersal from rookeries east of 148W (including Ascension Island) account for approximately 70–90% of the particles reaching South American FGs, and approximately 5–65% of particles reaching Caribbean and North American FGs (see the electronic supplementary material, tables S3 and S4). This mixing was also noted in genetic studies, where the initially postulated east –west rookery differentiation [28] was not upheld when additional data were analysed [12]. By contrast, west-to-east transport accounts for less than 9% of the particles reaching Mediterranean and African FGs, with the exception of western rookeries accounting for 97 –99% of particles reaching the Canary Islands (see the electronic supplementary material, tables S3 and S4). These simulations are in agreement with north–south differentiation of haplotype lineage distributions, with mixing in central areas [10]. South Atlantic rookeries account for approximately 2–30% of the particles reaching northwest Atlantic FGs, with highest contributions at FGs at the periphery of the Caribbean (e.g. Barbados; figure 2; electronic supplementary material, tables S1, S3 and S4). Within our models, rookeries in the northwest Atlantic do not contribute particles to FGs south of the
3
Proc R Soc B 280: 20131468
3. Results
similarly well correlated with most FG-centric MSAs, indicating the robustness of these methods.
rspb.royalsocietypublishing.org
simulations, a single value is produced for each rookery and FG, whereas MSA generates a probability with a mean and 95% CI. We used the mean MSA estimate, as is conventional in similar studies [16]. We did not ‘correct’ the particle tracking estimates to reflect the fewer sites available in MSA, because our goal was to compare the techniques with their full suite of strengths and limitations. We used Pearson’s correlation tests and the slope of the regression line to assess the relationship between particle tracking and MSA.
(a)
(b)
(c)
(d)
(e)
(f)
4
45° N
rspb.royalsocietypublishing.org
30° N 15° N 0° 15° S 30° S
Proc R Soc B 280: 20131468
45° S 45° N 30° N 15° N 0° 15° S 30° S 45° S 45° N 30° N 15° N 0° 15° S 30° S 45° S 90° W
0.4
60° W
1.2
30° W 0° density 2.0
30° E
60° E
90° W
60° W
30° W
0°
30° E
3
4
60° E
age 2.8
3.6
1
2
5
Figure 2. Predicted distribution of 13 500 virtual particles tracked forwards from (a,b) Costa Rica, (c,d ) Guinea Bissau and (e,f ) Ascension Island. (a,c,e) indicate particle density per grid cell throughout the simulations (counted daily, logarithmic scale, not weighted by population size). (b,d,f ) indicate the mean age (in years) of particles. Other conventions as in figure 1.
Equator (see figure 2a,b and the electronic supplementary material, table S1 and S2). Minimal dispersal is predicted from Mediterranean Sea rookeries (see figure 1 and the electronic supplementary material, tables S1, S3, S4 and figures S1, S2), consistent with genetic studies [11]. Our simulations also help to resolve ambiguous MSA results. Naro-Maciel et al. [10] concluded that m2m MSAs provided insufficient resolution to distinguish between contributions of Ascension Island and Guinea Bissau. Although m2m analyses indicate a substantial number of Guinea Bissau turtles forage in Brazil, o2m MSAs tend to assign those turtles to Ascension Island [10]. Our simulations support the o2m results; Guinea Bissau contributed no particles to southern FGs based on forward-tracking simulations (figure 2c) and backtracking suggests that only limited connectivity is likely (e.g. figure 3e). Our simulations predicted variable turtle age across the Atlantic, which in principle will be useful for validating
and refining our model (figure 1c). The data on missing age-classes will also be informative for managers to consider in conservation planning that requires knowledge of age structure, such as marine protected areas [16,17]. On average, older particles are found in the northeastern and central southern Atlantic, whereas younger particles are predicted throughout the Caribbean Sea, equatorial Atlantic and Mozambique Channel. Mean particle age is older in the coastal regions of the Gulf of Mexico, as well as the Bahamas and Bermuda where juveniles forage [29], but younger in deeper waters of the Gulf of Mexico and along the United States East coast. Additionally, gradients in age are predicted along the South American coast with mean particle age increasing from the Equator to the south. The older particle age along coastal areas of southern Brazil, Uruguay and Argentina, corresponds well with developmental habitats for green turtles, which are probably sites of first juvenile recruitment from oceanic habitats [30]. Furthermore, the
(a)
(b)
(c)
(d)
(e)
(f)
5
45° N
rspb.royalsocietypublishing.org
30° N 15° N 0° 15° S 30° S
Proc R Soc B 280: 20131468
45° S 45° N 30° N 15° N 0° 15° S 30° S 45° S 45° N 30° N 15° N 0° 15° S 30° S 45° S 90° W
0.4
60° W
1.2
30° W 0° density 2.0
30° E
60° E
90° W
60° W
30° W
0°
30° E
3
4
60° E
age 2.8
3.6
1
2
5
Figure 3. Predicted distribution of 21 900 virtual particles tracked backwards from (a,b) Barbados, (c,d ) Cape Verde Islands and (e,f ) Ubatuba, Brazil. (a,c,e) indicate the density of particles per grid cell throughout the simulations (counted daily, logarithmic scale), identifying connectivity ‘hot spots’ between oceanic locations and the FG. (b,d,f ) can be interpreted as the mean number of years a particle would have to drift from a particular location before reaching the FG.
oldest particles from Ascension and Trindad drift to the Indian Ocean, an exchange that is thought to occur only rarely in turtles [25,26]. For turtles to avoid this same fate, they may have evolved oriented swimming responses to promote retention near coastal FGs in lower South America. Comparing rookery-centric particle tracking results with those from rookery-centric MSA can provide insights into the relative roles of ocean currents and swimming behaviour on turtle distribution [9]. Because our model considers only movement due to surface circulation, it might best account for the distribution of turtles from rookeries, where swimming is a minor component of net velocity (i.e. current speed is high). For example, MSA and particle tracking were best correlated for Florida, USA (r ¼ 0.98, p , 3.8 1026), a rookery adjacent to the extremely fast-flowing and consistently directional Gulf Stream [20]. Additionally, the juveniles sampled at FGs may be considerably older than five years and have, presumably, shifted from behaviour
that promotes dispersal via ocean currents to ‘homing’. Turtles from the Suriname rookery, where particle and MSA results were not well correlated, are known to forage in northeastern Brazil [10,27,31]. These FGs are counter to predominant currents and transport there is probably achieved by older turtles swimming through coastal waters. The integration of high-resolution ocean circulation models and genetic MSA appears to be a powerful tool for generating and testing hypotheses regarding dispersal and distribution in the open sea [1,9,14]. This approach can be applied to numerous questions ranging from the direction and distance of dispersal in deep-sea invertebrate larvae [5,32], to the duration of drift between suitable habitats for rafting communities associated with macroalgae [33]. Given the limited information on organisms in the open sea (e.g. swimming behaviour, growth and mortality) [6], these models are perhaps most appropriately used as ‘null hypotheses’ of marine organism distributions that can guide the
Acknowledgements. We thank Philippe Verley for his development of freely available ICHTHYOP particle tracking software and Michael McDonald for maintaining freely available HYCOM output. Funding statement. Funding was provided by Oregon State University (N.F.P.) and College of Staten Island, City University of New York (E.N.M.).
References 1.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
turtles. J. Biogeogr. 37, 1752–1766. (doi:10.1111/j. 1365-2699.2010.02326.x) Bowen BW, Karl SA. 2007 Population genetics and phylogeography of sea turtles. Mol. Ecol. 16, 4886–4907. (doi:10.1111/j.1365-294X.2007.03542.x) Monza´n-Argu¨ello C, Dell’Amico F, Morinie`re P, Marco A, Lo´pez-Jurado LF, Hays GC, Scott R, Marsh R, Lee PLM. 2012 Lost at sea: genetic, oceanographic and meteorological evidence for storm-forced dispersal. J. R. Soc. Interface 9, 1725 –1732. (doi:10.1098/rsif.2011.0788) Proietti MC, Reisser JW, Kina PG, Kerr R, Monteiro DS, Marins LF, Secchi ER. 2012 Green turtle Chelonia mydas mixed stocks in the western South Atlantic, as revealed by mtDNA haplotypes and drifter trajectories. Mar. Ecol. Progr. Ser. 447, 195–209. (doi:10.3354/meps09477) Blumenthal JM et al. 2009 Turtle groups or turtle soup: dispersal patterns of hawksbill turtles in the Caribbean. Mol. Ecol. 18, 4841–4853. (doi:10.1111/ j.1365-294X.2009.04403.x) Godley BJ, Barbosa C, Bruford M, Broderick AC, Catry P, Coyne MS, Formia A, Hays GC, Witt MJ. 2010 Unravelling migratory connectivity in marine turtles using multiple methods. J. Appl. Ecol. 47, 769 –778. (doi:10.1111/j.1365-2664.2010.01817.x) Chassignet EP, Hurlburt HE, Smedstad OM, Halliwell GR, Hogan PJ, Wallcraft AJ, Baraille R, Bleck R. 2007 The HYCOM (Hybrid Coordinate Ocean Model) data assimilative system. J. Mar. Syst. 65, 60– 83. (doi:10.1016/j.jmarsys.2005.09.016) Lett C, Verley P, Mullon P, Parada C, Brochier T, Penven P, Blank B. 2008 A Lagrangian tool for modelling ichthyoplankton dynamics. Environ. Model. Softw. 23, 1210–1214. (doi:10.1016/j. envsoft.2008.02.005) Putman NF, He R. 2013 Tracking the long-distance dispersal of marine organisms: sensitivity to ocean model resolution. J. R. Soc. Interface 10, 20120979. (doi:10.1098/rsif.2012.0979) Putman NF, Shay TJ, Lohmann KJ. 2010 Is the geographic distribution of nesting in the Kemp’s ridley sea turtle shaped by the migratory needs of offspring? Integr. Comp. Biol. 50, 305– 314. (doi:10. 1093/icb/icq041) Putman NF, Scott R, Verley P, Marsh R, Hays GC. 2012 Natal site and offshore swimming influence fitness and long-distance ocean transport in young sea turtles. Mar. Biol. 159, 2117–2126. (doi:10. 1007/s00227-012-1995-5)
23. Lohmann KJ, Putman NF, Lohmann CMF. 2012 The magnetic map of hatchling loggerhead sea turtles. Curr. Opin. Neurobiol. 22, 336–342. (doi:10.1016/j. conb.2011.11.005) 24. Staaterman E, Paris CB, Helgers J. 2012 Orientation behavior in fish larvae: a missing piece to Hjort’s critical period hypothesis. J. Theor. Biol. 304, 188–196. (doi:10.1016/j.jtbi.2012.03.016) 25. Roberts MA, Schwartz TS, Karl SA. 2004 Global population genetic structure and male-mediated gene flow in the green sea turtle (Chelonia mydas): analysis of microsatellite loci. Genetics 166, 1857– 1870. (doi:10.1534/genetics.166.4.1857) 26. Bourjea J, Lapegue S, Gagnevin L, Broderick D, Mortimer JA, Ciccione S, Roos D, Taquet C, Grizel H. 2007 Phylogeography of the green turtle, Chelonia mydas, in the Southwest Indian Ocean. Mol. Ecol. 16, 175 –186. (doi:10.1111/j.1365-294X.2006. 03122.x) 27. Naro-Maciel E, Becker HJ, Lima EHSM, Marcovaldi MA, DeSall R. 2007 Testing dispersal hypotheses in foraging green turtles (Chelonia mydas) of Brazil. J. Hered. 98, 29 –39. (doi:10.1093/jhered/esl050) 28. Encalada SE, Lahanas PN, Bjorndal KA, Bolten AB, Miyamoto MM, Bowen BW. 1996 Phylogeography and population structure of the Atlantic and Mediterranean green turtle Chelonia mydas: a mitochondrial DNA control region sequence assessment. Mol. Ecol. 5, 473 –483. (doi:10.1111/j. 1365-294X.1996.tb00340.x) 29. Meylan PA, Meylan AB, Gray JA. 2011 The ecology and migrations of sea turtles. 8. Tests of the developmental habitat hypothesis. Bull. Am. Mus. Nat. Hist. 357, 1–70. (doi:10.1206/357.1) 30. Gallo BMG, Macedo S, Giffoni BB, Becker JH, Barata PCR. 2006 Sea turtle conservation in Ubatuba, southeastern Brazil, a feeding area with incidental capture in coastal fisheries. Chelonian Conser. Biol. 5, 93 –101. (doi:10.2744/1071-8443(2006)5[93: STCIUS]2.0.CO;2) 31. Schulz JP. 1975 Sea turtles nesting in Surinam. Zool. Verh. Leiden 143, 3–172. 32. Young CM et al. 2012 Dispersal of deep-sea larvae from the intra-American seas: simulations of trajectories using ocean models. Integr. Comp. Biol. 52, 483 –496. (doi:10.1093/icb/ics090) 33. Fraser CI, Nikula R, Waters JM. 2010 Oceanic rafting by a coastal community. Proc. R. Soc. B 278, 649–655. (doi:10.1098/rspb. 2010.1117)
Proc R Soc B 280: 20131468
Cowen RK, Gawarkiewicz G, Pineda J, Thorrold SR, Werner FE. 2007 Population connectivity in marine systems: an overview. Oceanography 20, 14 –21. (doi:10.5670/oceanog.2007.26) 2. Block BA et al. 2011 Tracking apex marine predator movements in a dynamic ocean. Nature 475, 86 –90. (doi:10.1038/nature10082) 3. Gaines SD, Gaylord B, Gerber LR, Hastings A, Kinlan BP. 2007 Connecting places: the ecological consequences of dispersal in the sea. Oceanography 20, 90 –99. (doi:10.5670/oceanog.2007.32) 4. Munk P et al. 2010 Oceanic fronts in the Sargasso Sea control the early life and drift of Atlantic eels. Proc. R. Soc. B 277, 3593 –3599. (doi:10.1098/rspb. 2010.0900) 5. McClain CR, Stegen JC, Hurlbert AH. 2011 Dispersal, environmental niches and oceanic-scale turnover in deep-sea bivalves. Proc. R. Soc. B 279, 1993 –2002. (doi:10.1098/rspb.2011.2166) 6. Mansfield KM, Putman NF. 2013 Oceanic habits and habitats: Caretta caretta. In Biology of sea turtles, vol. 3 (eds J Musick, KJ Lohmann, J Wyneken), pp. 189 –205. Boca Raton, FL: CRC Press. 7. Reich KJ, Bjorndal KA, Bolten AB. 2007 The ‘lost years’ of green turtles: using stable isotopes to study cryptic lifestages. Biol. Lett. 3, 712 –714. (doi:10.1098/rsbl.2007.0394) 8. Carr A. 1980 Some problems of sea turtle ecology. Am. Zool. 20, 489–498. 9. Putman NF, Verley P, Shay TJ, Lohmann KJ. 2012 Simulating transoceanic migrations of young loggerhead sea turtles: merging magnetic navigation behavior with an ocean circulation model. J. Exp. Biol. 215, 1863 –1870. (doi:10.1242/ jeb.067587) 10. Naro-Maciel E, Bonioli ACV, Marine M, Almeida AP, Baptistotte C, Bellini C, Marcovaldi MA, Santos AJB, Amato G. 2012 The interplay of homing and dispersal in green turtles: a focus on the southwestern Atlantic. J. Hered. 103, 792 –805. (doi:10.1093/jhered/ess068) 11. Bolker BM, Okuyama T, Bjorndal KA, Bolten AB. 2007 Incorporating multiple mixed stocks in mixed stock analysis: ‘many-to-many’ analyses. Mol. Ecol. 16, 685–695. (doi:10.1111/j.1365-294X.2006. 03161.x) 12. Monzon-Arguello C, Lopez-Jurado LF, Rico C, Marco A, Lopez P, Hays GC, Lee PLM. 2010 Evidence from genetic and Lagrangian drifter data for transatlantic transport of small juvenile green
6
rspb.royalsocietypublishing.org
interpretation of observations [20]. As additional information on the biology of oceanic species becomes available, such factors can be readily incorporated into the existing model framework (e.g. [1,9,22]), and thus refine the quantitative estimates for the ecological and evolutionary processes driven by organismal movements.
Nathan F. Putman1 and Eugenia Naro-Maciel2 1
Department of Fisheries and Wildlife, Oregon State University, 104 Nash Hall, Corvallis, OR 97330, USA Department of Biology, College of Staten Island, City University of New York, 2800 Victory Boulevard, Staten Island, NY 10314, USA 2
Research Cite this article: Putman NF, Naro-Maciel E. 2013 Finding the ‘lost years’ in green turtles: insights from ocean circulation models and genetic analysis. Proc R Soc B 280: 20131468. http://dx.doi.org/10.1098/rspb.2013.1468
Received: 7 June 2013 Accepted: 15 July 2013
Subject Areas: ecology, environmental science, genetics Keywords: dispersal, distribution, marine turtles, Chelonia mydas, ocean currents, population structure
Author for correspondence: Nathan F. Putman e-mail: [email protected]
Electronic supplementary material is available at http://dx.doi.org/10.1098/rspb.2013.1468 or via http://rspb.royalsocietypublishing.org.
Organismal movement is an essential component of ecological processes and connectivity among ecosystems. However, estimating connectivity and identifying corridors of movement are challenging in oceanic organisms such as young turtles that disperse into the open sea and remain largely unobserved during a period known as ‘the lost years’. Using predictions of transport within an ocean circulation model and data from published genetic analysis, we present to our knowledge, the first basin-scale hypothesis of distribution and connectivity among major rookeries and foraging grounds (FGs) of green turtles (Chelonia mydas) during their ‘lost years’. Simulations indicate that transatlantic dispersal is likely to be common and that recurrent connectivity between the southwestern Indian Ocean and the South Atlantic is possible. The predicted distribution of pelagic juvenile turtles suggests that many ‘lost years hotspots’ are presently unstudied and located outside protected areas. These models, therefore, provide new information on possible dispersal pathways that link nesting beaches with FGs. These pathways may be of exceptional conservation concern owing to their importance for sea turtles during a critical developmental period.
1. Introduction Perhaps one of the greatest obstacles in providing a scientific basis for conservation and management of marine species is the lack of basic information on their distribution across life-history stages [1]. For instance, many marine taxa have remarkable fidelity to specific reproductive areas, yet disperse widely over vast expanses of the ocean to forage [2]. Thus, processes that occur in one location may drive changes at distant sites or in other ecosystems through the ecological coupling of long-distance movement [3–5]. Gathering such information has proved difficult for species that spend much or all of their time in the open sea, where direct observation is hampered by logistical constraints [4–8]. This is particularly true of young green turtles (Chelonia mydas), which rapidly disperse from natal beaches into the ocean. Observations of these turtles are largely lacking during the subsequent life stage termed ‘the lost years’, and they are generally only seen again three or more years later upon recruiting to near shore foraging grounds (FGs) as juveniles [7,8]. Here, we used a modelling approach to predict the oceanic distribution of juvenile green turtles during their ‘lost years’ throughout the Atlantic Ocean, Mediterranean Sea and the southwestern Indian Ocean. We used complementary techniques in which thousands of virtual particles were released into an ocean circulation model in the vicinity of major rookeries and known FGs (figure 1a). Particles released near rookeries were tracked forwards through time to estimate likely dispersal routes, whereas particles released near FGs were tracked backwards through time to identify possible oceanic corridors available to turtles for reaching their FGs. By tracking passively drifting particles from their sources (rookeries) and endpoints (FGs), we bracket the range of likely dispersal pathways and diminish potential errors associated with our lack of information on post-hatchling swimming behaviour [9].
& 2013 The Author(s) Published by the Royal Society. All rights reserved.
30° N 15° N 0° 15° S
25 000 1800 200
30° S
2. Material and methods
30° N 15° N 3.6
15° S 30° S
density
0°
2.8 2.0 1.2 0.4
45° S (c) 45° N 30° N 15° N
5 4
15° S
3
30° S 45° S
age
0°
2 1
90° W
60° W
30° W
0°
30° E
60° E
Figure 1. Maps of model domain and predicted green turtle distribution. (a) Locations of FGs (green circles) and rookeries (white squares) examined in this study. Sizes of squares are scaled to relative rookery size; abundance at FGs is unknown. (b) Predicted distribution of particles released at rookeries in proportion to population size. Colours indicate particle density within a grid cell throughout the simulations (counted daily, logarithmic scale). Orange-red areas indicate possible ‘lost years hotspots’ where turtles might be found in high densities. (c) Mean age of particles (in years) within each grid cell (counted at daily intervals) throughout the 5-year simulations.
We compared our modelling results to estimates from published genetic mixed-stock analyses (MSAs) [10–12]. The ‘many-to-many’ (m2m) MSA method, an extension of the original ‘one-to-many’ (o2m) MSA, uniquely incorporates metapopulation structure of multiple FGs and rookeries through two approaches that correspond to the two particle tracking techniques described earlier [11–13]. In FG-centric MSAs, the proportion of turtles at a given FG, which comes from each source population (rookery), is estimated based on mitochondrial control region sequences. In rookery-centric MSAs, the proportion of turtles from a given rookery, which go to different FGs, is estimated. Confidence intervals (CIs) for MSA estimates are typically wide owing to inherent limitations: some locations are unavailable for sampling and
We tracked the dispersal of virtual particles throughout the Atlantic and southwestern Indian Oceans, and the Mediterranean Sea, within hindcast output from the Global Hybrid Coordinate Ocean Model (HYCOM) (http://hycom.org). HYCOM output has a spatial resolution of 0.088 (approx. 6 –9 km grid spacing), a daily time step, and assimilates satellite and in situ measurements to produce realistic velocity fields [18]. Numerical experiments were performed with ICHTHYOP (v. 2) particle tracking software [19]. For advection of particles, ICHTHYOP implemented a Runge–Kutta fourth-order time-stepping method whereby particle position was calculated each half-hour. To simulate sub-grid scale turbulent processes, horizontal dispersion was also included in the advection process [19]. The HYCOM-ICHTHYOP system accurately predicts the movement of surface-drifting buoys [20] and is well established for studying sea turtle dispersal [6,9,20–23]. Young turtles are relatively weak swimmers and divers, and although even minimal swimming can influence the distribution of marine organisms [9,24], these turtles’ net movement is largely driven by ocean currents [6,12,15]. Additionally, simulating swimming behaviour would introduce uncertainty that, without a priori information on orientation behaviour of different populations (e.g. [23]), would be impossible to parametrize. Likewise, we did not specify ‘recruitment’ (halting movement after a certain age or location encountered, e.g. [21]) to allow a full range of transport possibilities to be mapped. Dispersal was simulated from 28 rookeries (2.582.58 zones) in the Atlantic and southwestern Indian Ocean, and the Mediterranean Sea (see figure 1a and the electronic supplementary material, table S1). Fifty particles were released daily during the peak 90 days of each rookery’s hatching season in 2004, 2005 and 2006 (a total of 13 500 particles per rookery; electronic supplementary material, table S1). Each particle was tracked forwards through time for a total of 5 years. We recorded the percentage of particles that entered 32 FGs (58 58 zones) that are known to host green turtles (figure 1a). To estimate the distribution of pelagic juvenile green turtles, we plotted the number of particles within each grid cell throughout the 5-year simulations and the mean age of particles within each cell. The number of particles plotted was weighted by an estimate of population size (number of nesting turtles) at the source rookeries (see the electronic supplementary material, table S1). To further explore connectivity between rookeries and FGs, 100 particles were released every 5 days at the 32 FGs from the previous analysis. Releases occurred from 1 January 2012 through to 1 January 2009 (i.e. over 3 years for a total of 21 900 particles per FG). Particles were tracked backwards through time for a total of 5 years. Thus, ICHTHYOP determined where a particle came from to reach its final location in the FG. We recorded the percentage of particles that entered the 28 rookeries defined in the previous analysis. We weighted the particle tracking results based on rookery population size to determine the contribution of each rookery to a given FG. We determined whether results from particle tracking simulations were correlated with m2m MSA estimates from Naro-Maciel et al. [10] (i.e. 13 rookeries and 13 juvenile FGs). Because the source location of all particles is known in our
Proc R Soc B 280: 20131468
45° S (b) 45° N
2
rspb.royalsocietypublishing.org
there are often broadly shared haplotypes among rookeries [6,13]. Thus, considering population connectivity based on physical transport processes is probably necessary to help guide MSA interpretation [12,14–17]. Here, the combination of oceanic dispersal simulations and genetic MSA estimates offer a substantive hypothesis regarding basin-wide population connectivity and provide unprecedented information on the distribution of young turtles during their ‘lost years’.
(a) 45° N
(a) Rookery-centric estimates of connectivity Plotting the predicted distribution of virtual particles in proportion to rookery size indicated spatial variation in particle abundance and age structure across the Atlantic (figure 1). In the North Atlantic, a high abundance of particles was predicted from the southwestern Caribbean Sea through the Gulf of Mexico along the eastern United States coastline to North Carolina (approx. 378 N) and throughout the Sargasso Sea (approx. 20–358 N and 75–308 W). A more localized abundance of particles was predicted offshore of Guinea Bissau (near the west coast of Africa, approx. 128 N). In the Mediterranean, particle hotspots were restricted to the eastern basin. In the South Atlantic, high density was predicted between approximately 158 S–308 S and west of 308 W; to the southwest of this region a less dense cluster of particles was also predicted at approximately 308 S and 158 W. In the southwestern Indian Ocean, a substantial number of particles was predicted throughout the Mozambique Channel and just south of 308 S between 308 E and 758 E. The percentage of particles forward-tracked from rookeries to FGs was significantly correlated with the rookery-centric mean MSA estimates for five of the 13 rookeries (r . 0.74, p , 0.003) (see the electronic supplementary material, table S2).
(b) Foraging ground-centric estimates of connectivity The per cent contribution of rookeries to a particular FG was estimated by both backtracking particles from FGs and forward-tracking particles from rookeries (see the electronic supplementary material, tables S3 and S4). Simulations indicated that a high degree of connectivity is possible among rookeries and FGs. For example, backtracking simulations indicated that particles from 17 rookeries passed through the Barbados FG, with approximately 26% arriving from Suriname and approximately 22% from Guinea Bissau (rookeries on opposite sides of the Atlantic; electronic supplementary material, table S2). Where FGs are in close proximity to a large rookery, however, contributions may be dominated by a single source. For instance, the Nicaragua FG was predicted to have 85% of particles arrive from the Costa Rica rookery based on backtracking and 94% of particles estimated by forward-tracking simulations. The weighted per cent contributions of particles that were backtracked from FGs to rookeries were well correlated with FG-centric MSA estimates for nine of the 13 FGs (r . 0.59, p 0.032) (see the electronic supplementary material, table S5). The weighted per cent contributions of particles that were forward-tracked from rookeries to FGs were well correlated with FG-centric MSA estimates for 10 of 13 FGs (r . 0.57, p 0.033) (see the electronic supplementary material, table S6). Both particle tracking techniques were
4. Discussion Our findings provide critical predictive information on the ‘lost years’ of green turtles. Determining where marine turtles spend their first years of life has been a fundamental problem of sea turtle ecology for decades [6–8]. Our findings represent, to our knowledge, the first quantitative and testable hypothesis regarding the oceanic distribution and age structure of young green turtles throughout an entire ocean basin and beyond. Underscoring the biological and conservation significance of the results, this methodology can be used to inform demographic models that require spatially explicit and age-based information, and can be readily applied to other marine species. Most of the area covered by particles and many of the oceanic regions identified as possible ‘lost years hotspots’ for green turtles do not correspond to previously described FGs (figure 1b) and may be difficult to manage because they are outside of the Exclusive Economic Zone of sovereign nations. Further complicating management, simulations indicate that very different dispersal trajectories are likely for turtles departing from the same rookery (see figure 2 and the electronic supplementary material, table S1). For instance, particles dispersing from Ascension Island were transported to the North Atlantic, South Atlantic, and Indian Oceans (figure 2b). Likewise, backtracking simulations suggest that 26 of the 32 FGs could have inputs from rookeries across these three basins. Genetic surveys of southwestern Indian Ocean rookeries revealed patterns consistent with limited migration between the Atlantic and Indian basins [25,26], lending support to our findings. Mixing of diverse genetic stocks has also been documented in FGs where older turtles occur [10,13,27]. Thus, evidence increasingly suggests that the management of many turtle populations will require a broad approach across life-stages and regions [6,13]. Dispersal patterns (figures 1, 2 and 3) were consistent with previous genetic and oceanographic studies that suggest considerable connectivity is possible among green turtle populations during the ‘lost years’ [12,15,25–27]. Dispersal from rookeries east of 148W (including Ascension Island) account for approximately 70–90% of the particles reaching South American FGs, and approximately 5–65% of particles reaching Caribbean and North American FGs (see the electronic supplementary material, tables S3 and S4). This mixing was also noted in genetic studies, where the initially postulated east –west rookery differentiation [28] was not upheld when additional data were analysed [12]. By contrast, west-to-east transport accounts for less than 9% of the particles reaching Mediterranean and African FGs, with the exception of western rookeries accounting for 97 –99% of particles reaching the Canary Islands (see the electronic supplementary material, tables S3 and S4). These simulations are in agreement with north–south differentiation of haplotype lineage distributions, with mixing in central areas [10]. South Atlantic rookeries account for approximately 2–30% of the particles reaching northwest Atlantic FGs, with highest contributions at FGs at the periphery of the Caribbean (e.g. Barbados; figure 2; electronic supplementary material, tables S1, S3 and S4). Within our models, rookeries in the northwest Atlantic do not contribute particles to FGs south of the
3
Proc R Soc B 280: 20131468
3. Results
similarly well correlated with most FG-centric MSAs, indicating the robustness of these methods.
rspb.royalsocietypublishing.org
simulations, a single value is produced for each rookery and FG, whereas MSA generates a probability with a mean and 95% CI. We used the mean MSA estimate, as is conventional in similar studies [16]. We did not ‘correct’ the particle tracking estimates to reflect the fewer sites available in MSA, because our goal was to compare the techniques with their full suite of strengths and limitations. We used Pearson’s correlation tests and the slope of the regression line to assess the relationship between particle tracking and MSA.
(a)
(b)
(c)
(d)
(e)
(f)
4
45° N
rspb.royalsocietypublishing.org
30° N 15° N 0° 15° S 30° S
Proc R Soc B 280: 20131468
45° S 45° N 30° N 15° N 0° 15° S 30° S 45° S 45° N 30° N 15° N 0° 15° S 30° S 45° S 90° W
0.4
60° W
1.2
30° W 0° density 2.0
30° E
60° E
90° W
60° W
30° W
0°
30° E
3
4
60° E
age 2.8
3.6
1
2
5
Figure 2. Predicted distribution of 13 500 virtual particles tracked forwards from (a,b) Costa Rica, (c,d ) Guinea Bissau and (e,f ) Ascension Island. (a,c,e) indicate particle density per grid cell throughout the simulations (counted daily, logarithmic scale, not weighted by population size). (b,d,f ) indicate the mean age (in years) of particles. Other conventions as in figure 1.
Equator (see figure 2a,b and the electronic supplementary material, table S1 and S2). Minimal dispersal is predicted from Mediterranean Sea rookeries (see figure 1 and the electronic supplementary material, tables S1, S3, S4 and figures S1, S2), consistent with genetic studies [11]. Our simulations also help to resolve ambiguous MSA results. Naro-Maciel et al. [10] concluded that m2m MSAs provided insufficient resolution to distinguish between contributions of Ascension Island and Guinea Bissau. Although m2m analyses indicate a substantial number of Guinea Bissau turtles forage in Brazil, o2m MSAs tend to assign those turtles to Ascension Island [10]. Our simulations support the o2m results; Guinea Bissau contributed no particles to southern FGs based on forward-tracking simulations (figure 2c) and backtracking suggests that only limited connectivity is likely (e.g. figure 3e). Our simulations predicted variable turtle age across the Atlantic, which in principle will be useful for validating
and refining our model (figure 1c). The data on missing age-classes will also be informative for managers to consider in conservation planning that requires knowledge of age structure, such as marine protected areas [16,17]. On average, older particles are found in the northeastern and central southern Atlantic, whereas younger particles are predicted throughout the Caribbean Sea, equatorial Atlantic and Mozambique Channel. Mean particle age is older in the coastal regions of the Gulf of Mexico, as well as the Bahamas and Bermuda where juveniles forage [29], but younger in deeper waters of the Gulf of Mexico and along the United States East coast. Additionally, gradients in age are predicted along the South American coast with mean particle age increasing from the Equator to the south. The older particle age along coastal areas of southern Brazil, Uruguay and Argentina, corresponds well with developmental habitats for green turtles, which are probably sites of first juvenile recruitment from oceanic habitats [30]. Furthermore, the
(a)
(b)
(c)
(d)
(e)
(f)
5
45° N
rspb.royalsocietypublishing.org
30° N 15° N 0° 15° S 30° S
Proc R Soc B 280: 20131468
45° S 45° N 30° N 15° N 0° 15° S 30° S 45° S 45° N 30° N 15° N 0° 15° S 30° S 45° S 90° W
0.4
60° W
1.2
30° W 0° density 2.0
30° E
60° E
90° W
60° W
30° W
0°
30° E
3
4
60° E
age 2.8
3.6
1
2
5
Figure 3. Predicted distribution of 21 900 virtual particles tracked backwards from (a,b) Barbados, (c,d ) Cape Verde Islands and (e,f ) Ubatuba, Brazil. (a,c,e) indicate the density of particles per grid cell throughout the simulations (counted daily, logarithmic scale), identifying connectivity ‘hot spots’ between oceanic locations and the FG. (b,d,f ) can be interpreted as the mean number of years a particle would have to drift from a particular location before reaching the FG.
oldest particles from Ascension and Trindad drift to the Indian Ocean, an exchange that is thought to occur only rarely in turtles [25,26]. For turtles to avoid this same fate, they may have evolved oriented swimming responses to promote retention near coastal FGs in lower South America. Comparing rookery-centric particle tracking results with those from rookery-centric MSA can provide insights into the relative roles of ocean currents and swimming behaviour on turtle distribution [9]. Because our model considers only movement due to surface circulation, it might best account for the distribution of turtles from rookeries, where swimming is a minor component of net velocity (i.e. current speed is high). For example, MSA and particle tracking were best correlated for Florida, USA (r ¼ 0.98, p , 3.8 1026), a rookery adjacent to the extremely fast-flowing and consistently directional Gulf Stream [20]. Additionally, the juveniles sampled at FGs may be considerably older than five years and have, presumably, shifted from behaviour
that promotes dispersal via ocean currents to ‘homing’. Turtles from the Suriname rookery, where particle and MSA results were not well correlated, are known to forage in northeastern Brazil [10,27,31]. These FGs are counter to predominant currents and transport there is probably achieved by older turtles swimming through coastal waters. The integration of high-resolution ocean circulation models and genetic MSA appears to be a powerful tool for generating and testing hypotheses regarding dispersal and distribution in the open sea [1,9,14]. This approach can be applied to numerous questions ranging from the direction and distance of dispersal in deep-sea invertebrate larvae [5,32], to the duration of drift between suitable habitats for rafting communities associated with macroalgae [33]. Given the limited information on organisms in the open sea (e.g. swimming behaviour, growth and mortality) [6], these models are perhaps most appropriately used as ‘null hypotheses’ of marine organism distributions that can guide the
Acknowledgements. We thank Philippe Verley for his development of freely available ICHTHYOP particle tracking software and Michael McDonald for maintaining freely available HYCOM output. Funding statement. Funding was provided by Oregon State University (N.F.P.) and College of Staten Island, City University of New York (E.N.M.).
References 1.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
turtles. J. Biogeogr. 37, 1752–1766. (doi:10.1111/j. 1365-2699.2010.02326.x) Bowen BW, Karl SA. 2007 Population genetics and phylogeography of sea turtles. Mol. Ecol. 16, 4886–4907. (doi:10.1111/j.1365-294X.2007.03542.x) Monza´n-Argu¨ello C, Dell’Amico F, Morinie`re P, Marco A, Lo´pez-Jurado LF, Hays GC, Scott R, Marsh R, Lee PLM. 2012 Lost at sea: genetic, oceanographic and meteorological evidence for storm-forced dispersal. J. R. Soc. Interface 9, 1725 –1732. (doi:10.1098/rsif.2011.0788) Proietti MC, Reisser JW, Kina PG, Kerr R, Monteiro DS, Marins LF, Secchi ER. 2012 Green turtle Chelonia mydas mixed stocks in the western South Atlantic, as revealed by mtDNA haplotypes and drifter trajectories. Mar. Ecol. Progr. Ser. 447, 195–209. (doi:10.3354/meps09477) Blumenthal JM et al. 2009 Turtle groups or turtle soup: dispersal patterns of hawksbill turtles in the Caribbean. Mol. Ecol. 18, 4841–4853. (doi:10.1111/ j.1365-294X.2009.04403.x) Godley BJ, Barbosa C, Bruford M, Broderick AC, Catry P, Coyne MS, Formia A, Hays GC, Witt MJ. 2010 Unravelling migratory connectivity in marine turtles using multiple methods. J. Appl. Ecol. 47, 769 –778. (doi:10.1111/j.1365-2664.2010.01817.x) Chassignet EP, Hurlburt HE, Smedstad OM, Halliwell GR, Hogan PJ, Wallcraft AJ, Baraille R, Bleck R. 2007 The HYCOM (Hybrid Coordinate Ocean Model) data assimilative system. J. Mar. Syst. 65, 60– 83. (doi:10.1016/j.jmarsys.2005.09.016) Lett C, Verley P, Mullon P, Parada C, Brochier T, Penven P, Blank B. 2008 A Lagrangian tool for modelling ichthyoplankton dynamics. Environ. Model. Softw. 23, 1210–1214. (doi:10.1016/j. envsoft.2008.02.005) Putman NF, He R. 2013 Tracking the long-distance dispersal of marine organisms: sensitivity to ocean model resolution. J. R. Soc. Interface 10, 20120979. (doi:10.1098/rsif.2012.0979) Putman NF, Shay TJ, Lohmann KJ. 2010 Is the geographic distribution of nesting in the Kemp’s ridley sea turtle shaped by the migratory needs of offspring? Integr. Comp. Biol. 50, 305– 314. (doi:10. 1093/icb/icq041) Putman NF, Scott R, Verley P, Marsh R, Hays GC. 2012 Natal site and offshore swimming influence fitness and long-distance ocean transport in young sea turtles. Mar. Biol. 159, 2117–2126. (doi:10. 1007/s00227-012-1995-5)
23. Lohmann KJ, Putman NF, Lohmann CMF. 2012 The magnetic map of hatchling loggerhead sea turtles. Curr. Opin. Neurobiol. 22, 336–342. (doi:10.1016/j. conb.2011.11.005) 24. Staaterman E, Paris CB, Helgers J. 2012 Orientation behavior in fish larvae: a missing piece to Hjort’s critical period hypothesis. J. Theor. Biol. 304, 188–196. (doi:10.1016/j.jtbi.2012.03.016) 25. Roberts MA, Schwartz TS, Karl SA. 2004 Global population genetic structure and male-mediated gene flow in the green sea turtle (Chelonia mydas): analysis of microsatellite loci. Genetics 166, 1857– 1870. (doi:10.1534/genetics.166.4.1857) 26. Bourjea J, Lapegue S, Gagnevin L, Broderick D, Mortimer JA, Ciccione S, Roos D, Taquet C, Grizel H. 2007 Phylogeography of the green turtle, Chelonia mydas, in the Southwest Indian Ocean. Mol. Ecol. 16, 175 –186. (doi:10.1111/j.1365-294X.2006. 03122.x) 27. Naro-Maciel E, Becker HJ, Lima EHSM, Marcovaldi MA, DeSall R. 2007 Testing dispersal hypotheses in foraging green turtles (Chelonia mydas) of Brazil. J. Hered. 98, 29 –39. (doi:10.1093/jhered/esl050) 28. Encalada SE, Lahanas PN, Bjorndal KA, Bolten AB, Miyamoto MM, Bowen BW. 1996 Phylogeography and population structure of the Atlantic and Mediterranean green turtle Chelonia mydas: a mitochondrial DNA control region sequence assessment. Mol. Ecol. 5, 473 –483. (doi:10.1111/j. 1365-294X.1996.tb00340.x) 29. Meylan PA, Meylan AB, Gray JA. 2011 The ecology and migrations of sea turtles. 8. Tests of the developmental habitat hypothesis. Bull. Am. Mus. Nat. Hist. 357, 1–70. (doi:10.1206/357.1) 30. Gallo BMG, Macedo S, Giffoni BB, Becker JH, Barata PCR. 2006 Sea turtle conservation in Ubatuba, southeastern Brazil, a feeding area with incidental capture in coastal fisheries. Chelonian Conser. Biol. 5, 93 –101. (doi:10.2744/1071-8443(2006)5[93: STCIUS]2.0.CO;2) 31. Schulz JP. 1975 Sea turtles nesting in Surinam. Zool. Verh. Leiden 143, 3–172. 32. Young CM et al. 2012 Dispersal of deep-sea larvae from the intra-American seas: simulations of trajectories using ocean models. Integr. Comp. Biol. 52, 483 –496. (doi:10.1093/icb/ics090) 33. Fraser CI, Nikula R, Waters JM. 2010 Oceanic rafting by a coastal community. Proc. R. Soc. B 278, 649–655. (doi:10.1098/rspb. 2010.1117)
Proc R Soc B 280: 20131468
Cowen RK, Gawarkiewicz G, Pineda J, Thorrold SR, Werner FE. 2007 Population connectivity in marine systems: an overview. Oceanography 20, 14 –21. (doi:10.5670/oceanog.2007.26) 2. Block BA et al. 2011 Tracking apex marine predator movements in a dynamic ocean. Nature 475, 86 –90. (doi:10.1038/nature10082) 3. Gaines SD, Gaylord B, Gerber LR, Hastings A, Kinlan BP. 2007 Connecting places: the ecological consequences of dispersal in the sea. Oceanography 20, 90 –99. (doi:10.5670/oceanog.2007.32) 4. Munk P et al. 2010 Oceanic fronts in the Sargasso Sea control the early life and drift of Atlantic eels. Proc. R. Soc. B 277, 3593 –3599. (doi:10.1098/rspb. 2010.0900) 5. McClain CR, Stegen JC, Hurlbert AH. 2011 Dispersal, environmental niches and oceanic-scale turnover in deep-sea bivalves. Proc. R. Soc. B 279, 1993 –2002. (doi:10.1098/rspb.2011.2166) 6. Mansfield KM, Putman NF. 2013 Oceanic habits and habitats: Caretta caretta. In Biology of sea turtles, vol. 3 (eds J Musick, KJ Lohmann, J Wyneken), pp. 189 –205. Boca Raton, FL: CRC Press. 7. Reich KJ, Bjorndal KA, Bolten AB. 2007 The ‘lost years’ of green turtles: using stable isotopes to study cryptic lifestages. Biol. Lett. 3, 712 –714. (doi:10.1098/rsbl.2007.0394) 8. Carr A. 1980 Some problems of sea turtle ecology. Am. Zool. 20, 489–498. 9. Putman NF, Verley P, Shay TJ, Lohmann KJ. 2012 Simulating transoceanic migrations of young loggerhead sea turtles: merging magnetic navigation behavior with an ocean circulation model. J. Exp. Biol. 215, 1863 –1870. (doi:10.1242/ jeb.067587) 10. Naro-Maciel E, Bonioli ACV, Marine M, Almeida AP, Baptistotte C, Bellini C, Marcovaldi MA, Santos AJB, Amato G. 2012 The interplay of homing and dispersal in green turtles: a focus on the southwestern Atlantic. J. Hered. 103, 792 –805. (doi:10.1093/jhered/ess068) 11. Bolker BM, Okuyama T, Bjorndal KA, Bolten AB. 2007 Incorporating multiple mixed stocks in mixed stock analysis: ‘many-to-many’ analyses. Mol. Ecol. 16, 685–695. (doi:10.1111/j.1365-294X.2006. 03161.x) 12. Monzon-Arguello C, Lopez-Jurado LF, Rico C, Marco A, Lopez P, Hays GC, Lee PLM. 2010 Evidence from genetic and Lagrangian drifter data for transatlantic transport of small juvenile green
6
rspb.royalsocietypublishing.org
interpretation of observations [20]. As additional information on the biology of oceanic species becomes available, such factors can be readily incorporated into the existing model framework (e.g. [1,9,22]), and thus refine the quantitative estimates for the ecological and evolutionary processes driven by organismal movements.
