Animal tracking

All raw surface locations were filtered and regularized using a Bayesian State Space Switching Model (SSSM). Developed by [33], the SSSMs account for observation error, in order to regularize animal location estimates in time, as well as interpolate over small gaps resulting from missing observations of animals locations [1]. Final position estimates along each track were generated at 24-h intervals.

Characterization of basin-scale movements

In order to characterize long-term, large-scale movements of juvenile loggerheads in the North Pacific, we applied several filters to account for tagging bias and deploy location, and the presence of short or incomplete tracks [34, 35]. Tracks were first categorized by deploy region (Japan or CNP). Individuals that transmitted for less than 60 days were removed from the data set. From these data, spatial density maps were used to calculate areas of high residency, based on deploy location. Using hexagonal polygon binning, we calculated the number of days spent in a 1° longitudinal area, similar to [36] (Fig. 2). The movements of each individual track were classified into one of three dominant migratory routes, based on direction of travel. Throughout all tracks, the seasonal north-south migration with the Transition Zone Chlorophyll Front (TZCF) is evident, as described by [31]. Because we have long-term, multi-year data for these animals, we can go beyond seasonal movements to look at the larger, basin scale movements of their migrations. For this reason, movements were categorized by their east-west dynamics: (1) moving eastward within the CNP foraging grounds (e.g. Fig. 3a), (2) moving westward within the CNP (e.g. Fig. 3b), (3) moving eastward but underwent a considerable change in direction (e.g. moving east then west, or ‘turning around’, thereby staying within the CNP) (e.g. Fig. 3c), and (4) moving eastward to Baja California, Mexico foraging grounds (e.g. Fig. 3d). Since almost all tracks displayed portions of eastward migration (Fig. 2), tracks that moved east then west were identified by a subsequent longitudinal displacement of at least 3° in the westward direction (Fig. 2c).

Exploration of environmental and biological drivers of basin-scale movements

Track movements

Two hundred and thirty-one juvenile loggerhead sea turtles (23.3–83 SCL cm) were tracked from January 1997 – November 2013. One-hundred and ten aquarium-reared and 24 wild-caught individuals were released off of Japan. Seventy-four aquarium-reared and 23 individuals were released within the CNP. Deployments ranged from 41 to 1434 days (mean 351 days ± 256 days SD). Of these, 95 individuals transmitted for greater than 1 year; 14 for more than 2 years and 8 for more than 3 years.

Individuals deployed off Japan utilized the Kuroshio Extension Current (KEC) to disperse into pelagic areas (Fig. 2a). Individuals deployed within the CNP showed high use between 180° and 160°W (Fig. 2b). Mapping of loggerhead movements showed that irrespective of deploy location, turtles displayed an extended residence time (greater than 100 turtle days per grid cell), between 165°E–158°W longitudes (Fig. 2c). This long-term residency was demonstrated by several individuals that traversed back and forth within the CNP for several years (Fig. 4).

One hundred and forty-seven individuals (63.6 % of the total) displayed an eastward only migratory pattern before end of transmission (376.2 days ± 252.3 days), moving an average distance of 2028 km (±2454.3 km). Of these, ten were wild-caught turtles.

Nineteen sea turtles (8.2 % of the total), all wild-caught, moved westward from their CNP deploy locations, towards Japan (mean 199.6 days ± 192.9 days, 3105 km ± 2096.3 km). Sixty-five individuals (28.1 % of the total) moved eastward and then reversed direction along their migration routes, staying within the CNP (582 days ± 304.7 days, mean 11,128 km ± 4939.3 km). Five were wild-caught individuals. Thirty-four of these 65 turtles were deployed off of Japan and reached an average maximum eastward longitude of 180° before reversing direction (Additional file 1: Figure S1). The remaining 31 turtles were deployed throughout the Central North Pacific and reached an average maximum longitude of 160°W (Additional file 1: Figure S1). Only one out of all 231 turtles migrated to Baja California (Fig. 3d).

Environmental data

Median sea surface temperatures experienced by turtles was 17.6 °C (±2.2 °C, range of 10.0–28.74 °C) and 0.2 °C (±0.2 °C, range of 0–1.9 °C) for SST RMS. Median concentration of chlorophyll-a was 0.3 mg m^-3 (±1.8 mg m^-3, range of 0.0–72.8 log mg m^-3). Median magnetic field values were 51.3° (±5.1°, range of 26.3–62.1°) for inclination, -0.1° (±5.0°, range of -16.1–6.3°) for declination, and 43,820 nT (±1749.2 nT, range of 35,640–49,410 nT) for total intensity. Sea turtles deployed off of Japan experienced the greatest change in the Earth’s magnetic field declination, corresponding to 160° E - 180° (Fig. 5b-c). Individuals that were deployed within the CNP were deployed eastward of this gradient, and thus did not experience the same regional change in the Earth’s Magnetic declination as the individuals deployed off of Japan.

GAM results

No real-time environmental variables were significant in understanding a change in direction for the 65 individuals that moved from east to west (Additional file 1: Table S3). The two variables that were primarily attributed to a change in direction for the western Pacific deployed turtles were magnetic field declination (Fig. 5 and the number of days traveled (Additional file 1: Figure S4a-b)). For the central North Pacific, the two variables most correlated to a change in direction were magnetic field inclination and the month. These models revealed that western Pacific deployed turtles are more likely to continue traveling eastward under lower magnetic field declination values and for the first 250–300 days of travel, whereas the CNP turtles are more likely to continue traveling eastward under higher values of magnetic field inclination and towards the latter half of the year (Fig. 5 and Additional file 1: Figure S4c-d).

Hypothesis 1: CNP juveniles are not mature enough to recruit to BCP

To date, most studies of foraging loggerheads off the Baja California coast have focused on large juveniles (55–85 SCL cm) [28, 45]. For this reason, it could be suggested that turtles from this data set were too young to undergo an ontogenetic shift to neritic waters. However, recent skeletochronology has aged BCP turtles as young at 3 years old, which overlaps with ages of turtles tracked in this study [46]. This matches the age and size range of the one turtle from this data set that did migrate to the BCP. It should be noted that this turtle was originally deployed in the Central North Pacific, essentially giving it a head start towards its eastern foraging grounds. However, [28] found there to be no significant difference in the SCL sizes of CNP and BCP juveniles.

Hypothesis 2: Some utilize pelagic waters throughout their entire juvenile life history stage

Because loggerheads have been shown to take advantage of both oceanic and neritic habitats as both juveniles and subadults [47, 48], it is entirely possible that not all loggerhead turtles in the North Pacific undergo trans-Pacific dispersals, but instead use pelagic habitat for their entire juvenile phase, as suggested by [13]. A recent study of a subset of individuals from this population showed active orientation of juveniles within the CNP [22]. For this reason, it may be that the CNP is not just a migratory corridor that juveniles pass through on the way to foraging grounds off Baja California, but representative of juvenile foraging grounds altogether. The extended residence time within the CNP and lack of a migration to the BCP could be indicative of an alternative life history strategy for juveniles of this population [28].

Hypothesis 3: Turtles returning to previously experienced preferable habitat

For many migratory species (birds, tuna, sharks, and other species of sea turtles), complete migrations are not simple or direct [34, 49–51]. In fact, indirect routes are ‘followed not to only avoid unfavorable areas but as a pragmatic solution to completing a long journey successfully’ [50]. This may be especially true for early stage sea turtles, as they are free from the constraints of breeding, and are able to seek out the most productive areas to optimize growth while avoiding thermal stressors [28, 52]. Long distance migration across the CNP is likely energetically costly. Upon leaving the seasonally productive waters of the Kuroshio Extension Bifurcation Region (KEBR) and the Transition Zone Chlorophyll Front (TZCF) (see [13, 31], foraging opportunities in the eastern CNP may be increasingly difficult.

That turtles frequently appear to turn around as they move further east across the North Pacific and after a mean of 334.5 (±227.8 SD) days at sea, may be due to reaching a maximum energetic threshold (see [50]). Such a detour, or re-entry, of juveniles into the biologically productive waters of the KEBR and TZCF would allow turtles to minimize their energetic costs of travel, that prevents fasting and allows turtles to refuel for a potentially long-distance migration to the eastern Pacific developmental grounds. Similar behavior has been shown for green turtles (Chelonia mydas) in the South Atlantic, as they undergo transoceanic migrations between breeding sites (Ascension Island) and coastal foraging sites along Brazil [50].

Just how loggerheads may be able to successfully navigate a return to foraging hotspots has been better understood in other populations. In the North Atlantic Ocean, extensive research has shown that early stage loggerheads display a versatile navigational system. Namely, their open ocean migration is guided in part by passive drift associated with North Atlantic Subtropical Gyre (NASG) and Gulf Stream circulation, and active orientation due to swimming [9, 47, 53, 54]. There is a strong selective pressure for a juvenile turtle to remain within geographic areas that provide suitable conditions [55]. Poor navigation outside the gyre could lead to lethal temperatures [41]. This is done by active orientation in relation to a regional magnetic map [39–42, 54, 56]. These studies have effectively shown that the transoceanic migration of loggerheads in the North Atlantic, and the geographic regions they utilize along their migratory path, including ontogenetic shifts in habitat—are in large part bounded by the navigational markers associated with changes in the magnetic field [40–42]. In the North Pacific, magnetic influences have not been studied as in depth, however [25] noted that changes in the magnetic field could influence habitat choice on a basin-scale level.

Results from this study show that, similar to North Atlantic studies, North Pacific juveniles may use the Earth’s magnetic field [41] to reorient themselves back to the favorable habitat of the KEBR, while staying within thermally suitable latitudes. The isoclines of magnetic field inclination are similar to the latitudinal changes in SST (Fig. 5a and Additional file 1: Figure S3a). As individuals move with the north-south trend of the TZCF [20, 31], isoclines of inclination may help prevent animals from being swept by currents into inhospitable temperatures, also similar to the North Atlantic [41]. Turtles deployed off Japan travel through the Kuroshio Extension Current and Bifurcation Region (KEC and KEBR, respectively), and thus, a sharp gradient in declination as they make their way through the CNP, from 160°E - 180° longitude (Fig. 5b-c). This is the same region known for its biological productivity [13] (Additional file 1: Figure S3b). Several studies have examined the potential for magnetic maps to be ‘imprinted’ upon turtles throughout their transoceanic migrations [43, 57, 58]. Therefore, one explanation for a reversal in migration is the use of indirect movements based on the number of days traveled and the increase in energetic costs. As turtles move eastward throughout the CNP, foraging opportunities may be less prevalent, triggering them to turn around and return to more favorable habitat using a combination of regional navigational markers.

Hypothesis 4: Migration routes may be tied to genetics

Until the 1990s, the origins of Baja California loggerheads were entirely unknown. Work by [10] connected the haplotypes of Baja juveniles to observed haplotypes of loggerheads found off Japan. It is now known that nesting for the entire North Pacific loggerhead population is restricted to the Japanese Archipelago [10, 17, 59].

More recent work has begun to highlight significant differentiation among Japanese rookeries [11, 60] and their influences on the spatial distribution of feeding aggregations, suggesting that genetic composition of loggerheads may be tied to post-nesting migration patterns [61]. We hypothesize that these genetic differences may also express themselves during developmental migrations, such that there may be a genetic component from some nesting beaches that contribute to Baja California migrations. Despite the large sample size, it is possible that satellite tagged animals from this study representative of nesting beaches that do not display this genetic contribution. However, until there is better resolution of the genetics of juvenile North Pacific loggerheads, this hypothesis will remain difficult to elucidate.

Hypothesis 5: Environmental conditions influence recruitment to BCP

Like all ectotherms, sea turtles are inherently tied to the temperature of their surrounding environment. Studies have shown that higher water temperatures are energetically more favorable for sea turtles in terms of growth, digestion, and maintenance of core body temperature, up to a 30 °C thermal maxima [14, 62, 63]. Moreover, several management strategies currently use SST and ENSO events as a metric of bycatch avoidance for the North Pacific loggerhead population (see [24, 64]). During El Nino years, i.e. when SSTs are anomalously warmer, the California Drift Gillnet Fishery is closed due to increased interaction with loggerhead sea turtles [64]; an event that is not experienced during other oceanographic regimes. It is possible that interannual variability and/or anomalous SSTs play a larger role in east-west movements, promoting or prohibiting movement eastwards, towards Baja California, Mexico. However, more data is needed to fully explore this hypothesis.

Caveats

It should be noted that there are some caveats to these data, which may hinder our ability to fully understand the relationship between loggerhead movement and their environment. Daily SST measurements may capture mesoscale features, but may not be representative of relationships to anomalous events or larger-scale oceanographic and atmospheric variability (e.g. ENSO). The absence of a prey field and intermittent satellite data due to cloud cover (i.e. chlorophyll concentration) may prevent the model from being more robust. Ultimately, models that incorporate an index of forage and that address turtle energetics may be needed to further advance our understanding of movement. Further, the majority of the tracks represent captive-reared turtles (n = 204 out of 231 turtles). While some caution must be taken in comparing captive with wild caught turtle behavior, no study to date has observed significant differences in migration and swimming behavior between the two [13, 21]. However, further studies are needed to compare the habitat and behaviors of captive-reared and wild-caught sea turtles under similar environmental conditions [21].