1. Age at maturity is hard to estimate for species that cannot be directly marked or observed throughout their lives and yet is a key demographic parameter that is needed to assess the conservation status of endangered species.
2. For loggerhead turtles (Caretta caretta) in the North Atlantic and North Pacific, juvenile growth rates (c. 10 cm year−1) were calculated by examining size increases during transoceanic journeys; durations of which were estimated from satellite-tracked Lagrangian surface drifter buoy trajectories.
3. Lagrangian-derived growth estimates were used in a weighted loglinear model of size-specific growth rates for loggerhead turtles and combined with newly available information on size at maturity to estimate an age at maturity of 45 years (older than past estimates).
4. By examining the age at maturity for 79 reptile species, we show that loggerhead turtles, along with other large-bodied Testudine (turtle and tortoise) species, take longer to reach maturity than other reptile species of comparable sizes. This finding heightens concern over the future sustainability of turtle populations. By maturing at an old age, sea turtles will be less resilient to anthropogenic mortality than previously suspected.
Age at maturity has long been recognized as a fundamental life-history component integral to demographic models and hence assessments of the conservation status of species (Lewontin 1965; Berner & Blanckenhorn 2007). In general, traits such as age at maturity typically scale with body size and species exhibiting late age at maturity tend to show a number of k life-history characteristics such as low intrinsic rates of population increase (Saether & Bakke 2000; Cardillo et al. 2005; Nee et al. 2005; Depczynski & Bellwood 2006; Hard et al. 2008). For example, for northern right whales (Eubalaena glacialis) in the Atlantic, the late age at maturity contributes to low reproductive output, and hence, the premature death of even a few mature females (e.g. through ship collisions) may drive the population to extinction (Fujiwara & Caswell 2001). Obtaining accurate estimates of age at maturity is thus crucial for determining the potential susceptibility of species and populations to declines or extinctions and for effective conservation management.
While the age of maturity is easily determined for captive animals, such studies are limited in that they may not reflect the age at maturity for free-living individuals. A variety of approaches have thus been developed for estimating the age at maturity for free-living animals. For species that can be directly observed and identified throughout their lives (e.g. through photo-identification or tagging), it is possible to directly measure age at first breeding. When species cannot be tracked throughout their lives, it may be possible to estimate the age of individuals through a number of indirect ways. For example, developing fish can be aged based on growth rings in their otoliths, which have been used to show how commercial harvesting of fish has driven down the age and size of first breeding (Trippel 1995; Shackell et al. 2010). For other taxa such as mammals and reptiles, age can be estimated through skeletochronology using growth rings in structures such as teeth and bones (Amano et al. 2011). One of the limitations of this approach however is that these bony structures are susceptible to the loss of growth layers resulting in a tendency to underestimate age at maturity (Parham & Zug 1997). Other techniques include the analysis of length frequency distributions; however, limitations include assumptions that length frequency modes (if evident) represent single age classes (Bjorndal, Bolten & Martins 2000). Consequently, as a result of limitations and the intrinsic uncertainties of indirect ageing methods, age at maturity is still not well known for many species. This is particularly apparent for long lived and highly mobile marine species such as sea turtles (Cheloniidae and Dermochelyidae families) because of their life histories and associated logistical constraints, which prevents reliable and routine age assessments.
Directly recording the age at first breeding of free-living sea turtles is almost impossible because of difficulties associated with marking hatchlings in any way that they can be recognized later in life. While mark–recapture studies enable size-specific estimates of the growth rates of sea turtles at adult nesting and subadult coastal development habitats, the early pelagic life-history stage of post-hatchlings coupled with their small size and low survival probabilities have prevented any success with mark–recapture studies (Lutz & Musick 1997). Consequently, age at maturity in sea turtles has been estimated in a number of indirect ways (e.g. skeletochronology and length frequency analysis), which has contributed to the range of age at maturity estimates both between and within different loggerhead turtle populations. For example, estimates of the mean age at maturity for loggerhead sea turtles (Caretta caretta) in the western North Atlantic range from 6·5 years (Caldwell 1962) to 39 years (NMFS-SEFSC 2001). With the onset of better data and methods driving this upwards revision in age at maturity estimates, robustness of estimates is still hindered by the paucity of growth rate estimates for free-living post-hatchlings (Metcalfe & Monaghan 2001; Roark, Bjorndal & Bolten 2009).
To estimate post-hatchling growth rates, we estimated the time taken for individuals to drift from start (nesting) locations to end (stranding locations and/or development habitats) and the difference between the sizes of individuals between these two locations. Drift times were estimated using Lagrangian drifter trajectories, i.e. satellite-tracked drogues carried by the prevailing surface currents. We focused our analysis on loggerhead sea turtles in the North Atlantic and North Pacific because of the availability of Lagrangian drifter data and size records of individuals. We have previously advocated that in certain circumstances, it may be possible to use drift times of hatchling turtles from nesting beaches to known foraging sites to estimate growth (Hays & Marsh 1997), i.e. if hatchlings enter the water at size x and arrive elsewhere at size y after a time interval t, then their growth rate is simply (y − x)/t. With the advent of easily accessible global data sets for satellite-tracked Lagrangian drifter buoys (see Fossette et al. 2010; Hays et al. 2010) coupled with better sea turtle data sets on size distribution at different locations, it should now be possible to calculate robust estimates for post-hatchling growth rates. Here, we conduct this analysis for loggerhead sea turtles in two ocean basins and revise estimates for the age at maturity. We then perform a meta-analysis of reptilian age at maturity estimates to examine whether our revised estimates for sea turtles fit within general patterns across taxa. Finally, we discuss the conservation implications of our findings.
Materials and methods
Estimating Drift Times
All analyses were conducted using ESRI ArcGIS™ version 9.2 and R version 2.10.1 software (R Development Core Team 2009). Lagrangian drifter data were downloaded from the NOAA-AOML website: (http://www.aoml.noaa.gov/envids/gld/) with no restrictions on date or drogue attachment imposed. This data set spans from 1979 to present and contains quality controlled data of over 14,500 satellite-tracked surface buoys (drogued at 15 m to reduce wind effects) interpolated to provide fixes at 6-h intervals (Lumpkin & Pazos 2006). Owing to the proximity of the strong boundary currents to the major loggerhead rookeries in Florida and Japan (Gulf Stream and Kuroshio current, respectively) and the weak swimming ability of neonate turtles (Salmon & Wyneken 1987), only drifters passing within 50 km from major rookeries on the SE coast of USA and Japan were used to best represent drift trajectories that would be encountered by post-swimming frenzy hatchlings. Drift times for buoys that passed within 50 km of nesting beaches and reached within 50 km of locations for which size data of small loggerhead turtles were available were then calculated.
As a result of the long distance between nesting and re-encounter locations (several 1000 kms), very few buoys completed the entire drift. To estimate drift times between Florida-Northern Europe and Florida-Azores, we therefore split each journey into two legs. These legs were selected based on the routes buoys followed between the respective sites. For the Florida-Northern Europe journey, leg one ended in an area off Nova Scotia (delineated by longitudes: 55°W to 45°W and latitudes: 45°N′ to 35°N). For the Florida-Azores journey, leg one ended when buoys reached a longitude of 55°W (Fig. 1a). In the North Pacific, the first leg of the journey was defined when buoys first reached the oceanic area (delineated by longitudes: 166°E to 170°W and latitudes: 45°N′ to 28°N) where turtle size data were available. Owing to the time buoys spent entrained in eddies in this general area, the second leg of the journey was defined as the time buoys spent in this area before continuing along trajectories in the North Pacific gyre (Fig. 1b).
The impacts of any potential active directional swimming by post-hatchlings on drift times were estimated by placing realistic swimming behaviours (parameterized from laboratory experiments; Salmon & Wyneken 1987; Witherington 1991; Lohmann et al. 2001) using R software into passive particle trajectories from the surface current fields of a state-of-the-art ocean model hindcast. Passive particle trajectories were generated from particle releases from the SE USA using ARIANE software from an ocean model with 1/4° horizontal resolution based on the Nucleus for European Modeling of the Ocean (NEMO) (see Appendix S1 and Appendix S2, Supporting information).
Size of Loggerhead Turtles
Information on the size and locations of small turtles in the North Atlantic and North Pacific was determined from a variety of fishing records and stranding reports (see Table S1 and Appendix S2, Supporting information). In the North Atlantic, we compared the size of neonate hatchlings reported at major rookeries in Florida (Dodd 1988) with the size of post-hatchlings recorded around the Azores and in northern Europe (around the UK and France). In the Pacific, we compared the size of loggerhead hatchlings (Dodd 1988), with the size of post-hatchlings reported in the central North Pacific. Different criteria were used to define the size of small turtles at arrival because of data availability and biological considerations associated with the different sites (further discussed in the results).
Growth/Age at Size Curves
A comprehensive meta-analysis of size-specific loggerhead turtle growth rates was undertaken (see Table S2 and Appendix S2). Only direct measures of growth from marked and recaptured turtles were used to revise parameterization of loggerhead turtle growth curves. To assign body size to reported growth increments, we took arithmetic means of marked and recaptured carapace lengths and straight carapace lengths (SCL) were converted to curved carapace lengths (CCL) (Witherington 2002). Data from marked individuals were only included when intervals between marking and recapture were known to be ≥365 and ≤1460 days. When mean growth rates were reported for size classes of marked individuals, only mean values based on individuals with recapture intervals ≥365 days were included in analyses.
Mark–recapture intervals were controlled for time, as short intervals (days/months) fail to account for seasonality of growth and amplify human measurement error (Chaloupka & Musick 1997), while large recapture intervals/growth increments are not an accurate measure of size-specific growth as growth rates decline nonlinearly with increasing body size (Karkach 2006). Data availability dictated that our focal area for all subsequent analysis became the North Atlantic. Our post-hatchling growth rate estimate from the Azores was used to parameterize our growth curve because the growth rates of turtles stranding in Northern Europe are compromised due to cold stunning. Using direct measurements of size-specific growth supplemented with our size-specific growth rate estimate for post-hatchlings from the Azores, the relationship between body size and annual growth was modelled.
Three growth functions were tested, an exponential decay model, a loglinear model and a loglinear model weighted by the number of individuals used in each growth rate estimate. We would expect, a priori, that this weighted model would be the most appropriate approach when curve fitting because the literature-derived growth estimates were sometimes based on the growth rate of individual turtles, but sometimes mean values for up to 150 turtles in defined size classes were reported. Growth curves were constrained to not drop below 0·58 cm year−1; the most robust direct measurement of adult growth derived from mark–recapture data of 67 nesting adults (Bjorndal & Meylen 1983). Based on an initial hatchling size of 4·5 cm (Dodd 1988), functional curves between body size and age could then be derived from the modelled annual growth functions. Mean size of neophyte nesters from the SE USA (98·2 cm CCL; TEWG 2009) was used to derive mean age at maturity estimates, and mean size at recruitment from pelagic to coastal habitats (53 cm CCL; Bjorndal, Bolten & Martins 2000) was used to derive estimates of the mean length of the pelagic phase. Because of intensive tagging efforts at nesting rookeries in the SE USA, unmarked females (without tags or obvious tag scars) were inferred to be putative first time nesters as the probability of a female turtle not being detected on a well-monitored beach during her first nesting seasons is low.
Reptile Age at Maturity
All past age at maturity estimates for loggerheads turtle populations along with the most recent age at maturity estimates for other reptile species were drawn from the published literature (see Table S3 and Appendix S2). Data from both sexes were included by taking arithmetic means of age and straight carapace lengths (Testudines), published mass estimates (Dinosaurs) and snout-vent lengths (Squamates and Crocodilia). Lengths were then converted into mass based on species-specific length mass relationships when available else higher family/order level equations were used. Linear models were fitted to logged mass and age at maturity values for Testudine (turtle/tortoise) and Squamate (snake/lizard) species to investigate age and mass at maturity patterns between and within different reptile orders. All values were used to fit linear models including for the Testudines, the three negative outliers that were more in align with the data from the Squamates.
Drifter-Derived Growth Estimates
In the North Atlantic, a total of 96 buoys passed within 50 km from the East coast of the USA (Florida to North Carolina). Drift calculations to Northern Europe were based on 15 buoys, which completed leg 1 of this journey [drift time: 337 ± 266 days (mean ± SD)] and 34 buoys that completed leg 2 of this journey (439 ± 176 days). Drift calculations to the Azores were based on 22 buoys that completed leg 1 of this journey (301 ± 280 days) and 54 buoys that completed leg 2 of this journey (436 ± 387 days; Fig. 1a). In the North Pacific, a total of 385 buoys passed within 50 km of nesting beaches in Japan. Drift calculations to the area of high seas where turtle size data were derived were based on 83 buoys, which reached this area (leg 1; 320 ± 190 days) and the time these 83 buoys spent in this area before continuing along their trajectories in the North Pacific gyre (leg 2; 323 ± 265 days; Fig. 1b).
By combining all possible combinations of leg one and leg two drift times over 500 (15 leg one buoys × 34 leg two buoys), 1000 (22 × 54) and 6000 (83 × 83) possible drift time scenarios were calculated from Florida to the UK/France, Florida to the Azores and Japan to the central North Pacific, respectively. These calculations produced modal drift times for these three journeys of 650, 450 and 550 days, respectively (Fig. 1c). Based on the modal drift time to the Azores, adding 1, 2 and 3 h day−1 of directional swimming oriented by the geomagnetic field to the centre of the gyre (as described by Lohmann et al. 2001) lead to predicted increases in drift times of 45, 64 and 94 days, respectively.
Size of Loggerhead Turtles
Size data were available for a total of 273 loggerhead turtles encountered around the coast of the UK/France, 2318 individuals from the Azores and 52 from the Pacific. It was not always clear whether measurements were based on straight (SCL) or curved carapace lengths (CCL). However, we included all reported carapace lengths because our interest was on small turtles circa 20 cm in length and absolute differences between CCL and SCL of specimens is small and likely why distinctions between the two measurements are often not made (Witham & Futch 1977). In northern Europe, stranded loggerhead turtles ranged in size from 4·7 to 110·0 cm carapace length with a clear predominance for small turtles (≤25·0 cm, n = 244). For these turtles stranding in Northern Europe, the mid-point (20·0 cm) of the modal size class of small turtles was used to best represent size at arrival following direct drift from the SE USA (Fig. 2a).
The Azores and North Pacific are important pelagic development habitats in which turtles take up temporary residency upon arrival. The modal size class of small turtles encountered is thus not an accurate representation of size at first arrival as turtles spend protracted periods in these areas. For the Azores, the smallest turtles encountered were <15·0 cm (n = 9) and the first size class that turtles started to arrive in appreciable numbers was ≥15·0 to <20·0 cm (n = 150). Consequently, while the abundance of turtles peaked in the size class ≥20·0 to <25·0 cm (n = 405), we considered the size of first drifters to arrive from SE USA to be best represented by the mid-point (17·5 cm carapace length) of the smallest size class at which comparatively appreciable numbers of individuals were recorded (Fig. 2b). The size of turtles in the North Pacific ranged from 13·0 to 63·0 cm in carapace length. Within this size range, two size stages of turtles were apparent, those <30·0 cm carapace length (n = 8) and those ≥30·0 cm carapace length (n = 44). Mean size of the group of eight small turtles (19·0 cm) was used to best represent the size of the first turtles to drift from Japan to this high seas site (Fig. 2c). Based on mean size at hatching (Dodd 1988) and combining the increase in size of the individuals (between hatching and re-encounter) with the calculated modal time to travel between the respective sites, we estimate growth rates of post-hatchling loggerhead turtles travelling to northern Europe, the Azores and the central North Pacific of 8·7, 10·5 and 9·6 cm year−1, respectively (however, see Table S2, for the range of growth estimates based on different sizes at arrival and drift times).
Growth/Age at Size Curves
We found 27 published studies presenting growth rate estimates for loggerhead turtles; 21 of which were conducted in the North Atlantic/Mediterranean, one in the North Pacific and five in the South Pacific (Table S3). For the North Atlantic, six studies contained growth rate estimates with suitable mark–recapture intervals that enabled size-specific annual growth to be determined. Furthermore, Lagrangian estimates of drift times from Florida to the Azores (450 days) from which we derived growth rates ensured that problems with long growth intervals for parameterizing nonlinear growth functions were avoided.
Data from six marked individuals ranging in size at first capture from 26·0 cm (CCL) to 78·4 cm (CCL) came from turtles utilizing development habitats around the Azores. Data from seven marked individuals ranging in size at first capture from 55·0 to 83·4 cm (CCL) were recaptured in neritic habitats off the coast of North Carolina. Mean size and growth rate data based on 104 individuals spanning three size classes (from 50·0 to 79·0 cm CCL) off the coast of North Carolina were also available as was mean size and growth rate data from 67 mature turtles nesting in Florida. A chance stranding event of six alive or freshly dead post-hatchlings following a storm in the Azores prompted the only other estimate of post-hatchling growth since Hays & Marsh’s (1997) Lagrangian-derived estimate for cold stunned turtles stranding in Northern Europe. By adopting a similar approach to us, Bjorndal, Bolten & Martins (2000) used mean hatching date in the SE USA along with mean date of stranding to infer a drift time and hence growth rate estimate of 12 cm year−1 for turtles with a CCL of 7·3 cm.
The nonlinear relationship between growth rate and body size was modelled using data that encompass the full spectrum of loggerhead turtle size classes from hatchling (4·5 cm CCL) to mature adults (>98·2 cm CCL). Growth rates decreased markedly with increasing body size, and all three models produced a good fit to our data. The exponential decay model: growth rate (cm year−1 CCL) = 15·0 × exp (−CCL/36·0) − 0·38 (R2 = 0·94, F3,17 = 181, P < 0·001; Fig. S1a) and loglinear model: growth rate (cm year−1 CCL) = (−10·4 × log10 CCL) + 21·1 (R2 = 0·95, F1,18 = 310, P < 0·001; Fig. S1b) explained 94–95% of growth rate variance. However, the relationship was best described by a weighted loglinear model, which explained 98% of growth rate variance: growth rate (cm year−1 CCL) = (−10·6 × log10 CCL) + 21·5 (R2 = 0·99, F1,18 = 1701, P < 0·001; Fig. 3a).
Using these size-specific annual growth functions, we derived size at age curves. A broad range over the size at maturity was evident amongst neophyte nesters along the SE coast of USA [98·2 ± 5·5 (mean ± SD), range: 80·4–115·0 cm CCL, n = 826; TEWG 2009]; consequently, mean size at first maturity was used to derive mean age at maturity. Using this size, mean age at maturity is estimated at 42·4 years (exponential decay model: 95% CI 30·1–60·0; Fig. S3c), 44·1 years (loglinear model: 95% CI 36·2–52·5; Fig. S3d) and 45·0 years using our most robust (weighted loglinear model: 95% CI 38·0–52·0; Fig. 3b). These three similar estimates are all higher than previous estimates (see Fig. S3 and Table S4). Differences in age at maturity estimates derived from parameterization of loglinear weighed models with different growth rate estimates (incorporating directional swimming, different sizes at arrival and drift times) ranged from 43·0 to 47·0 years (Table S2).
Using a mean size at recruitment to coastal habitats of 53 cm CCL (Bjorndal, Bolten & Martins 2000), mean age at recruitment is estimated to be 7·6 years (exponential decay model: 95% CI 6·6–9·0; Fig. S3c), 7·7 years (loglinear model: 95% CI 6·0–8·8; Fig. S3d) and 8·0 years (weighted loglinear model: 95% CL 8·0–9·0; Fig. 3b), supporting the previous estimate of 8·2 years (Bjorndal, Bolten & Martins 2000). Confidence intervals around modelled size at age curves were smallest for juvenile and subadult size classes for all three models and increased as individuals reached mean size at maturity. For example, the ranges of age between upper and lower confidence limits were <1 year for all three models for turtles between 4·5 and 40 cm, and for our most robust weighted loglinear model (which had the lowest 95% CI range), this increased to <5 and 14 years for 80 and 100 cm CCL turtles, respectively (Fig. S2).
Reptile Age at Maturity
Age and mass at maturity data were obtained for a total of 43 Squamate species, 31 Testudine species and for five other large-bodied reptile species (two from the Crocodilia order and three from the superorder Dinosaurian) (Table S4). Squamate species ranged in mean body mass at maturity from the desert night lizard (Xantusia vigilis) at <0·1 g to the Komodo dragon (Varanus komodoensis) at 65·0 kg. Testudine species ranged in mean body mass at maturity from the loggerhead musk turtle (Sternotherus minor) at 0·1 kg to our loggerhead sea turtle estimate of 95·5 kg. The mass of other large extinct and extant reptile species ranged from the Mississippi alligator (Alligator mississippiensis) at 24·2 kg to the Tyrannosaurus rex at 2500·0 kg.
Across different reptile taxa, mean age at maturity tended to increase with mean mass at maturity. However, the form of this relationship differed between these two major reptile orders, revealing that turtle/tortoise species take longer to attain maturity than other reptiles of equal mass (Fig. 4). While higher than previous estimates, our revised age at maturity estimate for loggerhead turtles fits within the general size and age at maturity pattern observed within the Testudine order. This is further evidenced by excluding our loggerhead turtle estimate in the regression between size and age at maturity for Testudine species as our age of maturity estimate of 45 years fits within the predicted age range based on 95% confidence intervals around the regression for other Testudine species [mean: 28·0 years, 95% CI 15·0–50·0; log10 age (years) = 1·00 + (0·22 × log10 mass [kg]), F1,28 = 23·0, R2 = 0·45, P < 0·001].
It is well known that slow growth and late age at maturity are general attributes of reptiles and a consequence of their ectothermic nature (e.g. Nelson et al. 2002). Set against this backdrop, our results suggest that loggerhead sea turtles and Testudine species in general are older at maturity than Squamate species of equal mass. Our revised loggerhead turtle age at maturity estimate stems from the parameterization of our growth curve with new estimates of post-hatchling growth rates, published subadult and adult mark–recapture growth estimates (all with controlled mark–recapture intervals) and improved estimates of size at maturity. Consequently, we consider our best estimate of the mean age at maturity (45 years) to have been derived from the most robust loggerhead turtle growth curve to date and thus valuable for inclusion in across taxa meta-analysis of age at maturity estimates. Post-hatchling growth rates were inferred from Lagrangian estimates of drift times and size at arrival of small turtles at distant sites following direct drift from nesting sites. In so doing, we make two key assumptions, firstly that Lagrangian drifters can be considered to drift in the same way as post-hatchling sea turtles and secondly that we are able to accurately define size at arrival.
For post-hatchlings stranding in Northern Europe, data on both size and date of stranded individuals were available. Observed monthly frequencies of stranding events peaked in February, March and April, and expected monthly frequencies of stranding events (based on randomly generated normal distributions of Lagrangian drift times to the UK and mean day of hatchling) peaked in March, April and May. These broadly similar seasonalities support assumptions that post-hatchling drift times reflect passive Lagrangian drift. However, the slightly earlier seasonal peak in observed turtle stranding may reflect the fact that some post-hatchlings wash up earlier than expected based on passive drift alone following winter storms. Nevertheless, while hatchlings are known to drift passively with ocean currents (Witherington 2002), recent laboratory experiments have revealed that post-hatchlings are smart drifters that can orientate in response to the intensity and inclination of the geomagnetic field (Lohmann et al. 2001; Lohmann, Putman & Lohmann 2008; Putman et al. 2011).
The ability to orientate to the geomagnetic field is presumed to help drifting hatchlings remain in favourable development habitats within the North Atlantic Gyre because observed directional swimming is significantly orientated towards the centre of the ocean gyre. To date, the smallest turtles that have been successfully tracked are much larger (circa 300 g) than neonate turtles (circa 20 g). Because the details of this ‘neonate satellite tracking project’ are currently unpublished (see however perspectives article by Pennisi 2011), it has not yet been determined whether the movements of these small juveniles deviate from surface currents. However, visual inspection of the unpublished tracking data from this project (http://www.seaturtle.org/tracking/index.shtml?project_id=378) reveals that the tracks of these individuals seem to be in accordance with the trajectories of the Lagrangian drifters and passively drifting particles that we derived from a state-of-the-art ocean model hindcast to parameterize drift times. By placing realistic swimming behaviours observed from laboratory experiments into passive particle trajectories, we investigated the impacts that directional swimming may have on drift times and ultimately age at maturity estimates. When incorporating realistic directional swimming into growth rate estimates, our estimated age at maturity increased by just 1·0 years. Since active swimming in response to the geomagnetic field is not orientated towards or away from our end points, impacts of directional swimming on growth and age at maturity estimates appear to be limited, although our age at maturity estimates are considered conservative.
Along with estimating the age of post-hatchlings upon arrival at distant sites, which we accomplished using drift times (see above), the second key input to our post-hatchling growth rate calculation was estimating the size of hatchlings at arrival. Growth rate was then simply the size increase divided by age. For turtles found in northern Europe, there is a very clear modal size (20·0 cm). The most parsimonious explanation for this observation is that these post-hatchlings strand following direct drift from rookeries on the east coast of the USA (Hays & Marsh 1997), with the small numbers of larger juveniles expected to have arrived following more circuitous trips across the Atlantic. Around the Azores and in the North Pacific, turtles of a broad range of sizes are encountered, consistent with the view that upon arrival turtles remain at these sites which are utilized as foraging and developmental habitats. Hence, the smallest size at which comparatively large numbers of individuals were recorded was considered to best represent size at arrival. We are therefore confident that for each location we are using the best available estimates of size at arrival. Furthermore, our Lagrangian-derived growth rates support previous estimates of the age of small turtles in both the North Atlantic and North Pacific (Zug, Balazs & Wetherall 1995; Bjorndal, Bolten & Martins 2000) and help complete a general model that explains the vast majority (98%) of the variation in size-specific growth rates for loggerhead turtles. Nevertheless, for the Azores, a range of sizes at arrival and associated growth rate estimates were used to derive size at age curves producing a narrow range of mean age at maturity estimates of 43·0 to 47·0 years.
While much discussion remains over the suitability of different growth equations to derive growth and size at age curves of species (Day & Taylor 1997; Karkach 2006; Lee & Werning 2008), by using data from post-hatchlings to mature adults, we avoid the common practice of extrapolating models beyond empirical size ranges. In so doing, all three of our fitted growth curves provided similar mean age at maturity estimates adding to the robustness of our results. Furthermore, our size at age curves closely matched that of the von Bertalanffy growth curve conforming to the relationship found for other large-bodied turtle/tortoise species (Gaymer 1968; Frazer & Ehrhart 1985; Frazer, Gibbons & Greene 1990). The error associated with 95% CI around age at size estimates was lowest for the weighted loglinear model, demonstrating the utility of this modelling approach to derive age at maturity estimates. This error increases with increasing body size, an unavoidable property of a nonlinear growth model that approaches close to zero prior to maturity as opposed to a sign of weaker parameterization of the growth rates of large adults.
Age at maturity estimates are typically based on the mean size of breeding populations. However, for species with indeterminate growth patterns, it is necessary to discriminate between first time breeders and the rest of the nesting population to get reliable age and size at maturity estimates (e.g. Hawkes et al. 2005). Size at maturity can be highly variable for highly migratory and/or circumglobal species such as sea turtles because of variation in growth rates prior to maturity (Angilletta et al. 2004; Angilletta, Steury & Sears 2004; Schaefer & Walters 2010). Mean size of neophyte nesters from well-monitored beaches such as the SE USA thus provided the best indication of size at maturity and was used in our study.
Previous estimates of the mean age at which loggerhead turtles attain sexual maturity show a systematic shift from early estimates based on growth rates in captivity (<10 years) to a progressive upward revision of this estimate (to 39 years). This shift occurred as the slower growth rates of free-living animals became apparent through mark–recapture efforts. Our age at maturity estimates are thus in accordance with this trend and highlights the importance of the continual revision of age at maturity estimates as better data and techniques become available.
Furthermore, our revised estimate of 45 years based on a mean size of maturity of 98·2 cm CCL (TEWG 2009) supports the past highest mean age at maturity estimate of 39 years for North Atlantic populations (NMFS-SEFSC 2001) using a mean size at maturity of 97·2 cm CCL. This former estimate was derived from a growth curve using mark–recapture data of subadult turtles (≥50 SCL cm) that had recruited from pelagic to coastal habitats with recapture intervals of >0·9 years. Duration of the pre-recruitment pelagic stage was defined as 7 years based on length frequency analysis of pelagic turtles around the Azores and a chance storm driven stranding event of seven post-hatchling turtles (Bjorndal, Bolten & Martins 2000). Our Lagrangian drifter–derived post-hatchling growth estimate that stemmed from 1188 drift scenarios and 150 stranded post-hatchlings supplements the storm-derived post-hatchling growth estimate of Bjorndal, Bolten & Martins (2000) resulting in the most robust growth curve to date in terms of the parameterization of juvenile life stages. Furthermore, while subadult and adult growth rate estimates are now routinely derived from mark–recapture intervals of ≥0·9 year because of growth seasonality, we also included an upper limit of 4 years on mark–recapture intervals. Placing an upper limit on mark–recapture intervals should also be routine practice for species that exhibit nonlinear growth trajectories as size-specific growth rates are overestimated from long recapture intervals resulting in a tendency to underestimate age at maturity.
Inclusion of our loggerhead age at maturity estimate with that of 78 other reptile species revealed strong positive relationships between mean mass and mean age at maturity within the Squamate and Testudine orders. However, it is striking that the time taken for turtle/tortoise species to reach maturity is much longer than other large-bodied snakes, lizards, crocodiles and dinosaurs. While the reasons behind these differences are enigmatic, the finding of later age at maturity raises concerns over the future sustainability of turtle and tortoise populations, because age at maturity is one of the most important predictors of extinction risk for k-selected species where population trends are largely driven by survival rates of mature individuals (Gadgil & Bossert 1970; Heppell, Caswell & Crowder 2000; Saether & Bakke 2000; Fujiwara & Caswell 2001; Linares et al. 2007; Eldridge, Hard & Naish 2010).
For logistical reasons, monitoring efforts for threatened sea turtle species are directed at breeding grounds; this may, however, be a poor indicator of the status of populations because of the long maturation times highlighted by our study and the resultant time lag before population declines or changes such as climate altered sex ratios would be detectable in the size of the breeding stock (Davenport 1997; Wapstra et al. 2009; Witherington et al. 2009; Schwanz et al. 2010). In light of this ticking time bomb scenario, further precautions need to be built into management regimes. Our findings coupled with more comprehensive data sets and understanding of turtle life histories present a timely opportunity to revise growth curves and age at maturity estimates for other sea turtle populations using novel and more refined methodologies such as those advocated here. Furthermore, by highlighting the ease with which expanding Lagrangian data sets can be used to shed light on the ecology (in this instance growth rates) of species during life-history stages where more conventional approaches have struggled, it is hoped its use will have broader application amongst the wider marine community.
GCH conceived the project. RS assembled and analysed the data, RM ran the particle tracking model. RS and GCH wrote the manuscript. RS was funded through a PhD studentship awarded to GCH from the Natural Environment Research Council. Additional support was provided by the Climate Change Consortium for Wales (C3W). For helpful comments on the manuscript thanks also goes to Sabrina Fossette.
[Note added after online publication on 16 November 2011]
In the third paragraph of the Discussion, data and a website referred to are not the authors' own contribution. The authors regret the inclusion of these data without appropriate acknowledgement to the owners of the data.