Habitat use of adult Pacific bluefin tuna Thunnus orientalis during the spawning season in the Sea of Japan: evidence for a trade-off between thermal preference and reproductive activity

To examine the habitat usage of adult Pacific bluefin tuna (PBF), electronic tagging was conducted in the Sea of Japan during May and June of 2012−2017. Archival tags were internally im planted and pop-up satellite archival transmitting tags were deployed; data on the horizontal movements and diving behaviours of 36 individual PBF were successfully retrieved. In the summer spawning season, the tagged PBF were concentrated near Sado Island and Oki Island in the Sea of Japan, and they were distributed widely to the southwest (near Tsushima Island) or northeast (near the Tsugaru Strait) in the autumn and winter. We obtained the first long-term tracking record (246 d) for adult PBF, and this individual exhibited residency in a known spawning region during the spawning season in the proximity of warm-core eddy features. This fish spent most of the daytime below the thermocline between 30 and 150 m depths where the surface ambient temperature was 26.0 ± 1.5°C, but at night it ventured into the warm surface layer. Its whole-body heat transfer coefficient increased when it experienced warm waters (≥24°C), which we suggest is a physiological response to avoid overheating. The mean peritoneal cavity temperature was only 1.8°C higher than the ambient temperature, compared with 6.9°C higher during the cooler autumn−winter period. Our hypothesis is that the warm surface temperatures found in the spawning grounds induce a physiology−reproduction trade-off in adult PBF, which must behaviourally and physiologically therOPEN ACCESS Large Pacific bluefin tuna Thunnus orientalis. Photo: K. Fujioka moregulate their body temperature to gain spatial and temporal access to oceanographic conditions that may promote larval survivorship and growth.


INTRODUCTION
Understanding the reproductive ecology of Pacific bluefin tuna Thunnus orientalis (PBF) is essential to revealing the population dynamics of this commercially and ecologically important species. Recent ABSTRACT: To examine the habitat usage of adult Pacific bluefin tuna (PBF), electronic tagging was conducted in the Sea of Japan during May and June of 2012−2017. Archival tags were internally implanted and pop-up satellite archival transmitting tags were deployed; data on the horizontal movements and diving behaviours of 36 individual PBF were successfully retrieved. In the summer spawning season, the tagged PBF were concentrated near Sado Island and Oki Island in the Sea of Japan, and they were distributed widely to the southwest (near Tsushima Island) or northeast (near the Tsugaru Strait) in the autumn and winter. We obtained the first long-term tracking record (246 d) for adult PBF, and this individual exhibited residency in a known spawning region during the spawning season in the proximity of warm-core eddy features. This fish spent most of the daytime below the thermocline between 30 and 150 m depths where the surface ambient temperature was 26.0 ± 1.5°C, but at night it ventured into the warm surface layer. Its whole-body heat transfer coefficient increased when it experienced warm waters (≥ 24°C), which we suggest is a physiological response to avoid overheating. The mean peritoneal cavity temperature was only 1.8°C higher than the ambient temperature, compared with 6.9°C higher during the cooler autumn−winter period. Our hypothesis is that the warm surface temperatures found in the spawning grounds induce a physiology−reproduction trade-off in adult PBF, which must behaviourally and physiologically ther-research on PBF gonadal development and larval distribution has advanced our understanding of the spatial and temporal patterns of spawning activity (Chen et al. 2006, Tanaka et al. 2007, Ashida et al. 2015, Okochi et al. 2016. The Sea of Japan is one of the main spawning habitats for PBF, in addition to the waters near the Ryukyu Archipelago and Taiwan . Recent electronic tagging research and catch data have shown that mature PBF typically enter the Sea of Japan through the Tsugaru Strait from the Pacific Ocean and distribute southwards throughout the basin (Boustany et al. 2010, Fujioka et al. 2015, Tawa et al. 2017. Spawning likely occurs in offshore waters between Oki Island and the Noto Peninsula (36−39° N, 132−137° E) (Abe et al. 2014), as evidenced by higher catch rates of >130 cm fork length (FL) PBF (≥ 3 yr old PBF) which exhibit maturity rates > 92% during the summer months (Okochi et al. 2016, Ohshimo et al. 2018) and by the elevated abundance of PBF larvae observed in this area between June and August (Tanaka et al. 2007. However, the spatial and temporal patterns of the swimming behaviour of adult PBF in the spawning area have not been examined. Previous research conducted across tuna species supports the hypothesis that spawning activity is typically restricted to periods when sea surface temperatures (SSTs) range from 24−31°C (Schaefer 1998(Schaefer , 2001. Similarly, gonadal development and larval surveys conducted in the Sea of Japan support the idea that PBF spawning occurs in waters that are 24°C and warmer (Kitagawa et al. 1995, Tanaka et al. 2006, Abe et al. 2014, Okochi et al. 2016. It has been suggested that spawning activities occur under warmer SSTs as an evolutionary mechanism to increase larval survival and growth rates (Tanaka et al. 1996). Adult PBF are typically found in temperate waters between 14 and 19°C (Uda 1957), and the warm temperatures (≥ 24°C) experienced in the spawning grounds may be less than optimal metabolically (Blank et al. 2007). PBF have relatively high metabolic rates compared with other fish species, and they exhibit regional endothermy, meaning that they are capable of maintaining body temperatures up to 12°C warmer than ambient water temperatures (T a ) (Kitagawa et al. 2006a, Blank et al. 2007). Furthermore, Blank et al. (2007) conducted swim tunnel respirometry experiments on juvenile PBF that demonstrated a minimum metabolic demand in water temperatures between 15 and 20°C, and large-scale electronic tagging studies conducted in the Eastern Pacific showed that juvenile PBF typically occupy waters averaging 17°C . One short-term (12 d) electronic tagging record from a PBF (230 kg) near the Ryukyu Archipelago demonstrated a peak utilization of 26.2°C waters (ranging from 12.0−30.8°C), indicating that PBF experience exceptionally warm temperatures in their spawning habitats (Yamada et al. 2004). Unlike its other internal organs, the PBF heart operates at or near T a , and Blank (2004) indicated that prolonged exposure to temperatures near 30°C could lead to cardiac failure. In light of these previous studies, the warm SSTs (> 29°C during the summer) of the Sea of Japan spawning grounds may be optimal for larval survivorship and growth ); however, they may present a distinct physiological challenge to adult PBF.
The goals of this study were to identify (1) the seasonal horizontal distribution of adult PBF in the Sea of Japan and (2) the seasonal vertical distribution of adult PBF and the influence of water temperature on their vertical distribution. To answer these questions, an electronic tagging experiment including 2 types of electronic tags was conducted on adult PBF (≥ 3 yr old) in the Sea of Japan between 2012 and 2017. Additionally, we examined (3) the relationship between PBF physiological limitations and reproductive behaviours using high-resolution (5 s sampling interval) internally implanted archival-tag data (depth, T a , peritoneal cavity temperature [T b ], and geolocation). Such descriptive information is important for understanding the thermal physiology of PBF and the spatial and temporal patterns of their spawning behaviour.

Electronic tagging
To examine the habitat usage and swimming behaviour of adult PBF, an electronic tagging experiment was conducted off Sado Island in Niigata Prefecture in the Sea of Japan during May and June from 2012−2017. All PBF were caught in 2 commercially operated set nets off the northern coast of Sado Island (38. 23°N, 138.51°E and 38.31°N, 138.52°E) (see Fig. 1). PBF were transferred from the set net to a tagging cradle aboard the tagging vessel using a crane-operated landing net. Once the tuna was in the tagging cradle, a fabric eye cover was placed over its eyes, and a hose was inserted into its mouth to irrigate the gills with oxygenated seawater. Straight FL measurements were obtained from the tip of the rostrum to the fork of the caudal fin. Archival tags (LAT 2310: 16 mm diameter, 76 mm length, 45 g in air; LAT 2810L: 13 mm diameter, 44 mm length, 13 g in air; Lotek Wireless) were implanted into the peritoneal cavity using a sterile scalpel to make a small incision (2.0−2.5 cm long) through the ventral muscle wall that was subsequently closed with one suture (Ethicon suture, J365, 1, Vicryl undyed 27, CTX taper). The light level, T a , T b , and pressure data were archived in the tag memory every 5 s; the predicted tag lifetime was 2−3 yr. Additionally, pop-up satellite archival transmitting (PAT) tags (MiniPAT: 38 mm diameter, 124 mm length, 60 g in air; Wildlife Computers) were anchored into the dorsal musculature at the base of the second dorsal fin using a titanium dart and monofilament leader coated with Kevlar and silicone heat-shrink tubing. A secondary loop attachment utilizing a titanium dart and monofilament leader was inserted posterior to the tag to improve tag retention. Time-series data (T a and pressure) were recorded at 10 min intervals. Both archivaland PAT-tagged fish received a conventional tag (Hallprint) which was placed into the musculature at the base of the second dorsal fin, crossing the pterygiophores to ensure retention. The surgical procedures generally took between 30 and 120 s depending on the behaviour of the fish, and the tagged fish were released directly into the open ocean from the side of the tagging vessel located opposite to the set net.
A total of 30 archival-tagged PBF (mean ± SD: 138.7 ± 14.4 cm FL) and 33 PAT-tagged PBF (141.4 ± 15.5 cm FL) were released off Sado Island during May−June in 2012 and 2014−2017. Of these, 8 archival-tagged PBF were recaptured (26.7% of the total); however, 2 of these fish had shed their archival tags, and only the external conventional tags were recovered. Additionally, we were unable to successfully download data from one archival tag due to mechanical problems. Thirty-one of the PAT tags (94.0% of the total) were released to the surface before the scheduled pop-off period (8 mo). Three of the PAT tags were physically recovered, and their data were successfully downloaded. The percentage of successful data transmissions received via satellite ranged from 0.6−100% (mean: 87.6%) across all individuals. We examined the horizontal movements and vertical distributions recorded by both tag types and conducted a detailed analysis of the PBF's thermal physiology using fine-scale (5 s sampling rate) archival tag data. In particular, we focused on one individual (ID252) for which we obtained long-term (246 d), high-resolution data through the spawning season and described the seasonal changes in its behaviour. See Table 1 for the release and recovery details, summarized swimming depth, and T a and T b data for all individuals.

Estimation of fish geolocation
To estimate the daily positions of archival-tagged PBF, an integrated state-space Kalman filter statistical model was used to calculate geolocation errors, movement parameters, and the most likely tracks (Lam et al. 2008, Galuardi et al. 2010. Eight day satellite-derived composites of 0.1° gridded SST data from the Pathfinder Advanced Very High Resolution Radiometer (AVHRR) were used for this analysis and are available from the NOAA Bloom-Watch 360 website (http:// coastwatch. pfeg. noaa. gov/ coastwatch/ CW Browser WW 360. jsp). Geolocation estimates were further refined using bathymetric corrections that modified the daily maximum swimming depth relative to the regional sea-floor depth (ETOPO1) (Teo et al. 2007a, Galuardi et al. 2010). We ran a 128 scenario model by changing the initial values of the parameters for each individual tag record, and the highest likelihood scenario was ultimately selected as the most likely track (Fujioka et al. 2018a,b). In accordance with previous studies (Nielsen et al. 2006, Schaefer et al. 2011, the model parameters quantifying the northward and eastward movement components were set at 3−7 nautical miles d −1 , random errors in the longitude and latitude position estimates were set from 0.5−1.0° and 1−5°, respectively, and the smoothing radius of the SST data parameter ranged from 30−90 nautical miles. These estimates were obtained using an unscented Kalman filter (implemented with 'ukfsst' package version 0.3 in R; Lam et al. 2008), following the techniques described in Fujioka et al. (2018a,b).
We utilized a hidden Markov model (GPE3 software, Wildlife Computers) to estimate the daily positions of PAT-tagged PBF. The GPE3 model integrates the light level, SST, and bottom topography data (ETOPO1) to generate geolocation estimates that are accurate relative to real-time Argos-based track records (Braun et al. 2018). SST data were provided by the NOAA Office of Oceanic and Atmospheric Research Earth System Research Laboratory Physical Sciences Division (NOAA/OAR/ESRL PSD) (www. esrl. noaa. gov/ psd/). We ran 5 scenarios for each individual track record, changing the swimming speed parameter in 0.5 m s −1 increments from 0.5−3 m s −1 , and the scenario with the highest model score was selected as the most likely track. The GPE3 software estimated the location at various times each day (midnight, sunrise, noon, and sunset), but we utilized only the noontime estimates to maintain consistency with the method used to calculate the noontime location for the archival tags.

Analyses
The seasonal horizontal habitat preferences of PBF were investigated (summer: June−August; autumn: September−November; winter: December−February) by examining fixed kernel density outputs of the daily distribution of all tagged fish using non-parametric estimates (Kitagawa et al. 2006b, Boustany et al. 2010, Fujioka et al. 2018a. Kernel density estimates were obtained using R v.3.4.4 (The R Project for Statistical Computing: www.r-project.org/) with the MASS package (v.7.3-51.4). The search radius was set at 2 min degree −1 , and the smoothing parameters were estimated using the least-squares method. We examined the daily distribution of all tagged fish within the Sea of Japan, excluding one PAT-tagged individual (ID16P2015) that migrated out to the Pacific Ocean during the winter period (see Section 3.1 and Fig. S1 in the Supplement at www. int-res. com/ articles/ suppl/ m668 p001 _ supp. pdf).
We compared the oceanographic conditions in the known PBF spawning ground in the Sea of Japan, the most likely track (without geolocation errors) of one archival-tagged PBF (ID252) with high-resolution monthly satellite-derived SST (at a depth of 50 m) and model-derived current velocities (0.1° × 0.1°) from the Fisheries Research Agency (FRA-ROMS). A long-term (246 d) data set was successfully recorded for the tagged PBF, during which it moved to the known spawning area in the Sea of Japan during the spawning period (see Section 3.1). We calculated the daily mean surface T a and daily depth distributions recorded by the archival tag. The temperature-at-depth profile for each day was generated using a cubic smoothing spline and the mean swimming depth during the summer−winter period.
We proceeded to investigate patterns in PBF vertical habitat use in relation to thermocline depth across seasons within the Sea of Japan for all tagged PBF, and specifically for the individual PBF with the long-term record (ID252). Data from each of the archival tags (n = 4; except ID252) and PAT tags (n = 18; > 3 d) were aggregated and analysed. Although many of our tag deployments were short-term records, we observed clear trends in vertical distribution across individuals (see Section 3.2). We analysed the mean swimming depth within 10 m depth bins for both day and night based on the light intensity recorded by the tags, and we calculated the mean T a and T b within each depth bin. The thermocline depth was defined as the 10 m depth bin within which T a decreased by 0.8°C relative to the surface temperature (0−10 m depth) (Kara et al. 2000, Furukawa et al. 2017. Moreover, to examine the thermal preference of the PBF (ID252) in relation to the swimming depth in all seasons, we utilized a Spearman rank correlation analysis (p < 0.001) to compare the frequency of surface swimming (shallower than 10 m) and the PBF's thermal excess above the daily mean T a (T b − T a ) calculated for each day and night period.
Finally, we explored the whole-body heat transfer coefficient (k) of the archival-tagged PBF calculated during the spawning season (summer; surface T a ≥ 24°C) and non-spawning season (early summer and autumn−winter; surface T a < 24°C) to better understand how those tagged PBF (ID252, ID256, ID1723) responded to body temperatures near the theoretical upper thermal limit (> 30.0°C) during the spawning season. Electronic tag data were divided into hourly periods, and for each time window, we assumed that the k of the PBF was in a non-steadystate condition. The rate of change in body temperature, , can be described as follows: (1) where t is time and is the rate of change in body temperature due to metabolic heat production (Holland et al. 1992). We optimized the parameters k and for each time window in the same manner as Teo et al. (2007a). The T b for each period was modelled using Eq. (1), and we used non-linear least-square methods to estimate the values of k and that would provide the smallest sum of squares between the observed and predicted T b values (Optimization Toolbox, MATLAB R2018a; MathWorks). Kitagawa et al. (2006a) reported that estimated by behavioural measurement data fluctuates diurnally, and its value differs approximately 1.2− 2.1 times between day and night. Therefore, we set the initial values of the calculation parameters to those corresponding to the day and night to vary diurnally. First, we estimated the hourly prior for the whole-measurement data when k was fixed at 3.6 × 10 −4 s −1 , an arbitrary value. Then, we estimated the initial values of for day and night to determine the initial conditions for calculating the model parameters. To optimize the model, the initial conditions of in the day and in the night were used as the mean value in the day and mean value in the night, with as the initial condition; the hourly k was estimated simultaneously. To obtain reasonable values, we set the search range of within 2σ of the prior . The diel patterns (day and night-time) of the k values were statistically compared using a paired sample t-test.

Habitat usage
A total of 5 archival tags and 31 PAT tags were physically recovered or successfully transmitted data after being deployed on adult PBF (142.2 ± 15.3 cm FL). The duration of recorded data ranged from 7−246 d for the archival tags and 1−240 d for the PAT tags (Table 1) Fig. 1a. Fig. 1b,c demonstrate the variation in habit use between seasons in the Sea of Japan with latitude and longitude, respectively. All tagged PBF (n = 36) remained within in the waters near Sado Island after release in the early summer, and some individuals moved southwest towards the Noto Peninsula (n = 11) or northwards along the coast of Honshu (n = 11) (Fig. 1). In summer, tagged PBF were distributed between 2 areas: (1) offshore waters around Oki Island and (2) coastal waters between Niigata Prefecture and the Tsugaru Strait.
The first area (36−39°N, 132−137°E) overlaps with the well-known spawning area for PBF in the Sea of Japan (Abe et al. 2014, Okochi et al. 2016 and with a second area described in recent studies (Ohshimo et al. , 2018. In autumn, the area with the highest occurrence of PBF shifted to either the southwestern or northeastern edge of the Sea of Japan. During the winter, some of the tagged PBF dispersed southwards towards Tsushima Island and the East China Sea (n = 1) or through the Tsugaru Strait to the Pacific Ocean around the Shatsky Rise (30−40°N, 155− 165°E) (n = 1) (see Fig. S1 for full track records of PBF ID16P2015). We successfully recovered a long-term (246 d all daily location estimates that migrated to the known spawning area in the Sea of Japan during the summer spawning season (Fig. 2a). We closely examined this individual PBF's habitat use in relation to environmental parameters to provide new information on the spawning ecology of this species. As such, Fig. 2b,c shows the daily distribution of this PBF during August plotted with the satellite-derived subsurface temperature (at a 50 m depth) and surface current velocity and direction. This tagged PBF moved southward to the region around Oki Island before heading to offshore waters northwest of the island in the summer. This PBF's movements occurred in the vicinity of a warm water mass (> 20.0°C at a 50 m depth) within the centre of an eddy system derived from the Tsushima warm current. Subsequently, the PBF travelled throughout coastal waters west of Japan to wards Tsushima Island between autumn and winter.

Seasonal differences in vertical movements
Seasonal changes in surface T a appear to have influenced the vertical distribution of ID252 (Fig. 3). The daily mean surface T a experienced by the tagged PBF increased from 15.7°C after its release in May to 24.6°C by early July, and it further increased to 29.5°C by the middle of August. Subsequently, the surface T a rapidly decreased from 25.3−22.8°C by the middle of September and then decreased gradually to 13.4°C by the end of January (Fig. 3a). This tagged PBF spent 42.5% of the daytime at depths shallower than 10 m during the early summer, when the surface T a was below 24.0°C, and 20.9% during the autumn− winter months (Fig. 3b). Conversely, its utilization of the surface layer was notably lower (5.5%) during the summer months before and after the peak surface T a was recorded (Fig. 3b). The early summer and autumn−winter months in the central Sea of Japan are typified by vertical mixing of the water column and cooling of the surface layer, and under these conditions, this tagged PBF swam at mean (± SD) depths of 24.1 ± 13.4 and 36.5 ± 8.9 m, respectively (Fig. 3c). In contrast, a strong shallow thermocline develops during the summer, and this PBF swam deeper on average (48.9 ± 12.2 m; Steel-Dwass test, p < 0.01), just below the thermocline. Across the tag record, this individual adjusted its swimming depth in response to gradual seasonal shifts in thermocline depth.
The tagged PBF (ID252) exhibited a distinct shift in vertical distribution between day and night when it was in the vicinity of the Sea of Japan spawning grounds. Fig. 4 demonstrates a representative 6 d selection of time-series swimming behaviour (5 s sampling interval) plotted with T a and T b during the early summer (surface T a < 24°C; 31 May to 9 July) (Fig. 4a), summer (surface T a ≥ 24°C; 10 July to 16 September) (Fig. 4b), and autumn−winter (surface T a < 24°C; 17 September to 31 January) (Fig. 4c) periods. When the daily mean surface T a was below 24°C in the early summer and autumn−winter periods, this tagged PBF primarily swam near the surface and frequently dove deeper to 120 m depth at dawn, but the variability of T a with diving behaviour was low (Fig. 4a,c). Conversely, when surface T a exceeded 24°C in summer, this tagged PBF frequently dove to over 30 m and often remained at depth in cooler waters at approximately 15°C (T a ) for several hours during the daytime (Fig. 4b). During the night-time, this PBF dramatically increased its utilization of warm surface waters, and its T b increased significantly during these shallow excursions.
The time-series of swimming behaviours for the 4 other archival-tagged PBF are shown in Fig. 4d−i. In early summer, when the surface T a was below 24.0°C, the daily mean swimming depths of the tagged PBF were mostly shallow (ID260, 31.8 ± 31.3 m; ID256, 14.1 ± 19.6 m; ID259, 17.7 ± 23.4 m; ID1723, 11.0 ± 13.0 m) (Fig. 4d,e,g,h). We were able to track 2 of these individuals (ID256 and ID1723) through midsummer when they were distributed off the northern coast of Honshu. Although the daily mean surface T a rose above 24.0°C on 19 July 2012 (ID256) and 11 July 2017 (ID1723) after release, the T a s during the summer remained relatively low (23.9 ± 1.2 and 24.8 ± 1.3°C, respectively) compared with the T a s experienced by ID252 (27.1 ± 1.4°C) throughout the summer (Fig. 4f,i). On average, these 2 tagged PBF swam near the surface layer in the summer (ID256, 15.5 ± 30.5 m; ID1723, 11.2 ± 14.5 m), and they occasionally dove over 100 m into cool waters (<15°C) during the daytime. Summaries of the swimming depth, T a , and T b data for all individuals (including PAT-tagged PBF) are shown in Table 1.
We proceeded to investigate these seasonal and diel patterns in the PBF's habitat utilization more generally by examining its daytime/night-time swimming depth distribution alongside mean T a and T b profiles for each season (Fig. 5). During the early summer when the thermocline depth ranged from 40−50 m, the tagged PBF (ID252) spent most of its time in the mixed layer (from the surface to the thermocline depth) both during the daytime (79.6%) and at night (89.0%) (Fig. 5a). During this time, the mean T b recorded within the surface layer (0−10 m) was 25.0 ± 3.0°C, which was 4.6°C higher than the mean T a . The vertical mixing of the water column during respectively (Fig. 5b). This PBF spent little time in the mixed layer during the daytime (16.7% spent between 0 and 30 m); however, during the summer nights, it exhibited high utilization of the surface habitat and frequently ventured below the thermocline (< 30 m). This PBF spent 54.3% of the nighttime within the mixed layer and 45.7% of the nighttime below the thermocline during the summer. During this time, the mean T b recorded in the mixed layer was extremely high (28.8 ± 1.2°C at the surface), with a maximum T b record of 32.1°C. The mean T b recorded within the mixed layer was only 2.8°C warmer than the surface T a ; however, it was 10.2°C warmer than the T a on average when the PBF dove to cooler waters below the thermocline (16.2 ± 1.0°C at 100−110 m depth) (Fig. 5b). The vertical distributions of T a s and T b s for the archival-tagged PBF (not including ID252) and PATtagged PBF records (n = 18; > 3 d) are shown for each season in Fig. 5d−h. Overall, the PBF were distributed in the surface mixed layer (0−10 m depth) during the daytime (49.2−62.8%) and night-time (48.3−68.5%) in both the early summer (archivaltagged PBF, n = 4, Fig. 5d; PAT-tagged PBF, n = 18, Fig. 5f) and summer (archival-tagged PBF, n = 2, Fig. 5e; PAT-tagged PBF, n = 5, Fig. 5g). They experienced high T a (25.1 ± 0.9°C at a 0−10 m depth) throughout the summer season (Fig. 4f,i), while their mean T b was 26.5 ± 1.4°C at a 0−10 m depth (Fig. 5e). When the thermocline deepened to 40−50 m during the autumn−winter, the PBF (PAT-tagged PBF, n = 1, Fig. 5h) exhibited wider depth distributions, ranging between 0−40 m during the day (75.0%) and at night (73.5%).

Diel patterns of swimming behaviour and T a
To examine daily differences in the daytime and night-time swimming behaviours exhibited by the tagged PBF (ID252), we investigated the relationships between the utilization of depths shallower than 10 m and (1) the mean T a experienced by this PBF and (2) this PBF's mean T b elevation above the mean T a (Fig. 6). Across all seasons, this PBF markedly decreased its utilization of surface waters during the daytime as surface T a increased (Spearman rank correlation: r S = −0.70, n = 246 d, p < 0.001) (Fig. 6a). Furthermore, surface swimming behaviours (0−10 m) were rarely observed (3.3 ± 4.3%) when the surface T a exceeded 24.0°C during the summer. The PBF's thermal elevation above T a decreased as surface T a increased (r S = 12 Fig. 5. Frequency distributions of the swimming depth of Pacific bluefin tuna (ID252) between day and night with vertical profiles of mean (± SD) ambient water temperature (T a , blue) and peritoneal cavity temperature (T b , red) in (a) early summer (31 May to 9 July) (surface T a < 24°C), (b) summer (10 July to 16 September) (surface T a ≥ 24°C), and (c) autumn−winter (17 September to 31 January) (surface T a < 24°C). Aggregated data from (d,e) archival-tagged (n = 4) and (f−h) PAT-tagged (n = 18; > 3 d) fish are shown for each season −0.78, p < 0.001). These mean temperature differences were 4.6 ± 1.3°C in the early summer and 7.4 ± 2.0°C in the autumn−winter when the surface T a was below 24.0°C, while in summer it was only 1.5 ± 1.8°C (Fig. 6b). During the nighttime, we did not detect a significant correlation between the swimming depth and surface T a experienced by this PBF (r S = −0.18, p = 0.004); however, the utilization of the surface layer decreased as the surface T a approached 30.0°C (Fig. 6c). The mean frequency in the summer was higher than that in the daytime (18.7 ± 17.7%). The absolute difference between PBF T b and T a decreased significantly as the surface T a warmed across all night-time temperature records (r S = −0.74, p < 0.001), and the values were 4.8 ± 1.6°C in early summer and 6.9 ± 2.1°C in autumn−winter, compared with 1.8 ± 1.6°C in summer (Fig. 6d).

Whole-body heat transfer coefficient, k
The k and of the tagged PBF (ID252) were estimated using the heat budget model (Fig. 7). In the dT dt m Fig. 5. (continued) early summer season, the mean values of k were 4.91 × 10 −4 s −1 during the day and 3.80 × 10 −4 s −1 at night. In the summer season, the mean values of k were 4.28 × 10 −4 s −1 during the day and 6.10 × 10 −4 s −1 at night. In the autumn−winter season, the mean values of k were 3.62 × 10 −4 s −1 during the day and 3.41 × 10 −4 s −1 at night.
We observed a significant diel difference in the average k coefficient value for ID252 measured for every 12 h period during the summertime (paired samples t-test, t = −9.302, df = 66, p < 0.001) (Fig. 8).
These results demonstrate that individual ID252 exhibited behaviours in response to the diel cycle, 14 Fig. 6. Daily frequency at depths shallower than 10 m and temperature difference (peritoneal cavity temperature, T b , minus ambient water temperature, T a ) (°C) of the tagged Pacific bluefin tuna (PBF) (ID252) plotted with surface T a during (a,b) daytime and (c,d) night-time (n = 246 d). Spearman rank correlation test scores (r S ) are also shown. Vertical dotted lines: preferred temperature for PBF spawning (≥ 24°C) (Okochi et al. 2016 especially during the summer nights in the known spawning season in the Sea of Japan.

DISCUSSION
This is the first study to show long-term records (246 d) of swimming behaviour of an adult PBF (ID252) throughout the spawning period in the Sea of Japan. The aim of this study was to reveal the broad horizontal movements of adult PBF in the Sea of Japan using archival and PAT tags as well as to describe the fine-scale characteristics of habitat utilization in one of the important spawning areas for this species. We observed 2 dominant movement patterns within the Sea of Japan: one where tuna moved towards the centre of the Sea of Japan near Oki Island and another where they moved along the northern coast of Honshu between Niigata Prefecture and the Tsugaru Strait (Fig. 1). Both of these regions may represent spawning habitats for adult PBF (Okochi et al. 2016, as revealed by Okochi et al. (2016), who reported 95% maturity rates in PBF with an average FL of 133.6 cm captured in the Sea of Japan by purse seine fisheries. These previous results suggest that the PBF tagged in this study (average FL: 142.2 cm) were mostly mature. The region offshore of Oki Island is an especially well-documented spawning ground, as demonstrated by long-running surveys (1956−1989, 1979− 1988, 1999−2004, and 2009− 2015) of larval PBF abundance in the Sea of Japan (Tanaka et al. 2007, Abe et al. 2014. We note that the horizontal habitat usage results shown herein reflect a densityweighted bias around the tagging site resulting from the relatively shortterm tag recoveries/early PAT tag pop-offs. Further research is needed to clearly define the spawning habitat utilized by adult PBF in the Sea of Japan. During the autumn and winter period, adult PBF were dispersed widely across the Sea of Japan (near Tsushima Island or the Tsugaru Strait) (Fig. 1). One PAT-tagged PBF spent 6 mo in the Sea of Japan before migrating through the Tsugaru Strait in the winter to reach the vicinity of the Shatsky Rise (Fig. S1). This result suggests that some PBF may have a particular tendency to exploit distant foraging areas related to topographic features (Boustany et al. 2010, Fujioka et al. 2018b). Further research is needed to understand the movements of adult PBF within and outside of the Sea of Japan. 15 Fig. 7. Time-series data of (a) swimming depth, (b) ambient temperature with the low-pass value (T a ), along with observed and predicted peritoneal cavity temperature (T b ) (c) whole-body heat transfer coefficient (k), and (d) internal heat production (dT m ) of tagged PBF (ID252). The predicted values of peritoneal cavity temperature and k were estimated using an optimized heat budget model. Vertical grey bars: night-time Despite limited evidence stemming from one tagged PBF, our results suggest that oceanographic features and SSTs influence habitat selection by adult PBF during the spawning season. In summer, the tagged PBF (ID252) spent most of its time in the centre of the Sea of Japan, where warm eddy features persisted during the spawning season (Fig. 2). Similar eddy features have been shown to play an important role in the spawning ecology of Atlantic bluefin tuna Thunnus thynnus, and it is proposed that the oceanographical conditions within these structures support enhanced larval growth and survival (Teo et al. 2007b, Bakun 2013, Muhling et al. 2013). These dynamic current systems may also transport PBF larvae to warm nursery waters (Kitagawa et al. 2010, Fujioka et al. 2015 where they may become entrained in food-rich coastal habitats (Tanaka et al. 2007, Kodama et al. 2017) without being advected to cooler offshore waters. Indeed, in previous studies, oceanographic models have been generated based on larval thermal preference that simulate the horizontal dispersal of PBF larvae with good predictive capacity (Abe et al. 2014). Regarding the thermal conditions for PBF spawning, it is commonly assumed that adult tuna primarily spawn at SSTs above 24°C (Schaefer 2001, Okochi et al. 2016, and previous laboratory experiments suggest that 25°C is the optimal water temperature for PBF egg hatching (Miyashita et al. 2000). Warm temperatures likely provide metabolic benefits to PBF larvae and therefore enable faster growth during this key life stage (Tanaka et al. 1996). In the present study, the daily mean surface T a experienced by a tagged PBF (ID252) ranged from 24.6−29.5°C during the summer (Fig. 3). Therefore, these results corroborate the hypothesis that warm eddy features can be selected by adult PBF to provide optimal conditions for larval growth and survival. Only through continued tagging efforts and larval surveys will we be able to clearly understand the utilization of such oceanographic features for PBF spawning.
Our horizontal movement findings demonstrated that the adult PBF (ID252) selected warm-core eddy environments in the Sea of Japan; however, this tagged PBF consistently avoided these warm surface waters during the daytime by diving into deeper waters (Figs. 4b & 5b). When surface T a was less than 24.0°C (early summer and autumn− winter), the swimming depth of this tagged PBF was closely related to seasonal changes in the thermocline depth (Figs. 4a,c & 5a,c). Our results suggest that all the tagged PBF preferred to utilize the mixed layer when conditions were cooler (Figs. 4 & 5), a finding consistent with those of many previous PBF tagging studies (Kitagawa et al. 2000, 2004, Itoh et al. 2003, Furukawa et al. 2014. Conversely, when the surface T a rose above 24.0°C in the summer spawning season, the tagged PBF (ID252) frequently avoided warm surface water (mean: 26.0°C) and swam below the wide thermocline (30−40 m) during the daytime (Fig. 5b). Interestingly, during night-time, this PBF ventured into the warm mixed surface layer but also dove below the thermocline.
We observed an interesting discrepancy in the patterns of vertical habitat use between a subset of the tagged PBF. During the summer, some tagged PBF (ID256, ID1723, and ID16P2015) remained in the surface layer throughout the entire day (Figs. 4f,i & 5e,g) when distributed off the northern coast of Honshu in relatively warm surface waters (mean 24.4−25.1°C) in a shallow (10−20/20−30 m) thermocline environment. Conversely, the PBF (ID252) that visited the known spawning grounds in the summer months spent most of the daytime avoiding the warm surface layer (mean: 26.0°C). The more northerly distributed PBF demonstrated a significant tolerance for longterm exposure to high surface T a s relative to the PBF, but we found that the internal T b s of the northern group was noticeably cooler (mean: 26.5°C) relative to the T b of the PBF on the spawning grounds (mean: 28.8°C). This high average T b is close to the hypothesized upper thermal limit of 30°C, which we presume explains this individual PBF's need to behaviourally thermoregulate by remaining below the warm mixed layer.
Little is known regarding the diel timing of bluefin spawning activity; however, it may occur primarily in the evening between 17:00 and 22:00 h (Masuma 2006, Okochi et al. 2016. Our results showed that k was significantly higher at night during the summer spawning period (Figs. 7 & 8), which supported the capability of the PBF to adjust their physiological response to avoid overheating in the high-temperature environment found in the summer spawning habitat. Despite these high ambient temperatures, the persistent evening excursions into the warm surface layer detected in this study may represent the spawning activity of adult PBF. However, novel techniques and analyses are needed to determine when spawning actually occurs as it is often a spontaneous and short-lived event. Future analyses should explore the role of the lunar cycle, which may regulate spawning activity due to its influence on the predation risk for eggs and larvae during night-time (Shimose et al. 2018).
These consistent, prolonged evening excursions into the warm surface layer likely presented a physiological challenge to the adult PBF (ID252) (Fig. 6). Blank et al. (2007) studied juvenile PBF in a swim tunnel respirometer and established a U-shaped correlation between T a from 8−25°C and metabolic rate, with a minimum rate of oxygen consumption at 15°C (175 ± 29 mg kg −1 h −1 ). Regional endothermy has been well documented in PBF, which maintain higher body temperatures through the use of various counter-current heat exchange organs (Carey et al. 1971, Stevens et al. 1974, Kitagawa et al. 2006a). Kitagawa et al. (2006a) reported that the T b of adult PBF was on average 5−7°C higher than T a . The ability of PBF to warm their bodies may provide an advantage under cooler conditions; however, in the warm spawning grounds, endothermy may exacerbate the physiological challenge of spawning. Therefore, bluefin tuna seem to have an impressive adaptive capacity to handle high temperatures (Fig. 6) through their diving behaviour (Figs. 4 & 5) and thermal biology (Figs. 7 & 8). For example, when Atlantic bluefin tuna arrive at the warm spawning grounds (> 25°C) in the Gulf of Mexico, their k rapidly increase, possibly as a result of changes in vasodilation, cardiac output, and/or blood flow to the gills (Teo et al. 2007a, Graham & Dickson 2001, 2004. It is important to note that despite the ability of a PBF to elevate its T b , its heart operates at or near T a . Blank (2004) conducted laboratory experiments on juvenile PBF and reported that prolonged exposure to waters over 30°C could result in cardiac failure, especially when combined with the high activity levels associated with spawning.
The physiological effects of high SSTs experienced in the spawning grounds may be compounded by the kinetically demanding nature of PBF spawning activity. Based on a video from sea-cage aquaculture environments and personal communications, PBF spawning activity is associated with elevated swimming speeds as males aggressively pursue female mates (Gen 2015). Furthermore, relatively high mortality rates have been documented for Atlantic bluefin tuna caught during scientific longline surveys in the Gulf of Mexico, and it is plausible that the energetic demand of capture (in tandem with prolonged exposure to high SSTs) results in cardiac failure for these animals (Teo et al. 2007a).
In addition, an increased T b is one of the potential effects of feeding and digestion. We constructed a heat budget model to calculate internal heat production and k using all the T b data, including that obtained during feeding events. The estimated k during the summer was higher in the night-time than in the daytime (Fig. 8). PBF are visual, opportunistic predators, and most feeding activity occurs in the daytime (e.g. Kitagawa et al. 2004). Previous studies in PBF of specific dynamic action (elevation in metabolic activity associated with the ingestion, digestion, absorption, and assimilation of a meal) have shown that it takes several hours for the metabolic rate or T b to reach its peak and that T b remains elevated for 20−33 h after a meal (e.g. sardine Sardinops sagax) (Whitlock et al. 2013). Therefore, post-feeding heat production may extend into the night-time, and it is possible that k is actively adjusted to dissipate body 17 Fig. 8. Boxplots of diel differences in whole-body heat transfer coefficients (k) of tagged PBF (ID252, ID256, and ID1723) estimated from the heat budget model in the spawning season (summer). There was a significant diel difference in the mean of k for ID252 measured for every 12 h period during the summer months (paired samples t-test, t = −9.302, df = 66, p < 0.001). Horizontal line: median; top and bottom of boxes: third and first quartiles, respectively; whiskers: maximum and minimum; open circles: suspected outliers heat. Our results clearly imply that adult PBF are able to physiologically regulate their body temperatures in the warm surface layer at night. However, further research is needed to understand the complex physiological mechanisms of PBF and how these responses can be used to detect thermal signatures of spawning activity in this species.
The combined laboratory-and field-based evidence suggests that spawning-age PBF incur increased energetic demands and an increased risk of cardiac arrest in the warm spawning grounds. Adult PBF are clearly highly migratory and are more than capable of avoiding high-temperature water masses. Nonetheless, they appear to strike an ecology− reproduction compromise when they select oceanographic conditions that promote larval success. The results presented herein revealed the behavioural adaptations exhibited by PBF exhibit to survive warm evenings in the Sea of Japan spawning grounds: they remained deep in colder waters during the daytime and periodically dove to cooler waters below the surface layer at night.
In conclusion, we found that adult PBF utilized a wide horizontal range throughout the Sea of Japan. We were fortunate to document one tagged PBF (ID252) that migrated to the known spawning grounds during the suitable spawning period. This individual exhibited a clear diel pattern in vertical distribution, likely representing a behavioural thermoregulatory strategy to gain access to optimal spawning conditions. Further research is needed that draws upon a large number of fine-scale tag records to establish a more comprehensive biological and environmental survey of PBF spawning in the Sea of Japan. Such data will help to determine the breadth of the physiology−reproduction trade-off suggested by our archival tag data for adult PBF and potential ecological considerations (i.e. broad shifts in ocean circulation and SSTs). Such information will contribute to quantitative evaluations of the reproductive potential of PBF, facilitating improved management of this species.