We analyzed the Japanese LL logbook for 1996–2021 and the Japanese PL logbook for 1972–2020. These logbooks provide the most informative and comprehensive information regarding fishing operations because data submission is mandatory for vessels licensed by the Japanese government. They provide daily information on fishing operations, including dates, locations at noon (the resolution of the fishery location is 1 arc degree), sea surface temperature (SST, available since 2013 for PL), and catch weight. Other gear-specific information included the number of fish caught, hooks for LL, and poles for PL. ALB is a target species for both fisheries; however, these fisheries switch targets to other species depending on the availability of ALB (season and area). To focus on the ALB in subtropical and temperate areas in the northwestern and central Pacific Ocean (20°N<latitude<50°N, 130°E<longitude<210°E), we extracted records that had an ALB catch greater than 0. Unrealistic logbook records, such as those with extremely high catches and a low number of hooks (hooks<1,000) and poles (poles<3), were removed as outliers.

To compare the spatial patterns of fishery grounds and ALB hotspots between the LL and PL fisheries off Japan, we spatially calculated the catch per unit effort (CPUE) by dividing the catch by effort in a 1°×1° grid for each fishery. The efforts of LL and PL were defined as the number of hooks/1000 and poles, respectively. CPUEs were normalized by dividing by the largest value for each fishery, resulting in a normalized nominal CPUE to make a visual comparison between fisheries easier.

A series of tagging studies were conducted for ALB in three locations (around the coastal region: 26° to 30°N and 135° to 148°E, and off Japan: 30° to 34°N and 133° to 137°E, 40° to 42°N, and 154° to 155°E) from 2000 to 2004 with two types of archival tags. The tags used were LTD2310 (Lotek Wireless Inc., USA) and NMT v1.1 (NWT series version 1.1, Lotek Wireless Inc., USA), capable of measuring external and internal temperature, light intensity and pressure, which were set to record every 512 s.

Archival tags were surgically implanted into the peritoneal cavity of the captured ALB, and conventional tags (Hallprint Inc., Australia) with contact information were attached to the base of the dorsal fin. In total, 83 tagged ALB (fork length: 50–87 cm) were released, and 4 were recaptured (recapture rate: 4.8%, Table 1 ). Because migration paths have already been estimated using the Unscented Kalman filter-based geolocation method incorporating sea-surface temperature (Hokimoto and Kiyofuji, 2014), we referred to this published article for the estimated paths.

Our study aims to understand how the T_min limits the vertical behavior of ALB and their susceptibility to fishing operations on a daily scale, T_min is considered the temperature at which ALB can survive after being exposed for at least daytime. Minimum temperature based T_min estimates derived from tagged individuals typically occur when tuna dive beyond the thermocline. However, since they rapidly return to the warmer surface waters to thermoregulate (Holland et al., 1992; Kitagawa et al., 2001, 2004), such momental minimum records do not necessarily indicate the T_min that ALB can survive through a day. To address this issue, we calculated the daily averaged water temperature from four tagged ALB and used the lowest daily average temperature as the T_min from tagging data. Recognizing that a limited sample size (N=4) could overestimate T_min and we also analyzed comparing the temperature recorded on the fishery logbooks for PL and LL which provide a broader spatial and longer periods of the data coverage (See detail in Chapter 2.1) compared to tagging data. Although the temperature from the logbook is not completely comparable to the daily average temperature from tagging data as the temperature from the logbook is only available as a form of SST, it would be representative of the T_min when considering the surface vertical distribution of ALB. Consequently, we established the lower T_min using both datasets.

Vertical thermal profiles are representative of the ocean habitats utilized by tuna during their migration (Williams et al., 2015; Aoki et al., 2017). Temperature profiles collected with the electronic tags are used in a clustering analysis to characterize the ALB thermal habitat. Tagging data enabled us to directly reproduce the thermal profiles based on the wild behavior of tagged individuals. Here, a few steps were introduced to generate daily thermal profiles from the tagging data. The first step was to group the ambient temperature data for a day (24 hours) into 10 m depth bins from 0 to 200 m using corresponding depth data. Subsequently, the average temperature within each 10 m depth bin was calculated, which enabled us to provide a thermal profile from 0 to 200 m for a day. This calculation was applied to whole data periods from release to recapture. This step provided an entire dataset of the thermal profile for a specific day under the condition of the ALB, exploring a range of 0–200 m evenly. For datasets that did not sufficiently represent the vertical thermal profile for a day, the second step was applied. The second step was to deal with these missing values at each 10 m depth bin by interpolating them from adjacent days. This was achieved using the zoo package (ver 1.8.9) in R. Interpolation could have a negative impact on the clustering of the generated thermal profile; thus, we visually checked its effect by comparing the interpolated profile with the original profiles. The entire dataset was then used for clustering analysis.

Ideally, an entire dataset of thermal profiles from tag records detailing water temperature and swimming depth should be prepared. However, data at deeper depths (<50 m) tended to be scarce in the northern areas owing to the ALB behavior ( Figures 1B, C ; Supplementary Figure S1C ). To address this gap, we interpolated the depth temperature of missing data from adjacent days, which provided a complete profile; Since the interpolated data, representing 60% of the final dataset, could impact the clustering analysis, we checked the effect of interpolation by comparing each thermal profile with the original one. In all cases, both original and interpolated profiles visually agreed well with each other (see supporting results), confirming that the interpolated data accurately reflects the habitat characteristics pertinent to our clustering analysis.

We classified the daily thermal profiles into several clusters based on a common thermal habitat using the flexmix package in R (version 4.1.1) for model-based cluster analysis as a Gaussian mixture model (Leisch, 2004). This package also enabled us to assign the response and explanatory variables to the model. We used daily thermal profiles as multivariate response variables, assuming a Gaussian distribution, and applied the null model with only an intercept value of 1 with no explanatory variables. The thermal profile in our study area changes depending on spatial and temporal changes in the mixture of ocean currents (Kawai, 1972; Yasuda, 2003), implying potential explanatory variables of dates and locations (year, month, day, latitude, and longitude). However, a simple classification based solely on the thermal profile still has explanatory power for tuna ecology (Bestley et al., 2009; Aoki et al., 2017). Therefore, we kept the model simple rather than adding several explanatory variables. The decision on the number of clusters relied on Bayesian Information Criteria (BIC). We varied the potential number of clusters from 1 to 10, and the final number was determined from the number with the minimum BIC among the potential numbers.

A spatial comparison of CPUE between LL and PL fisheries revealed the geographical features of each fishery ( Figure 2 ). Overall, the LL fishery occurred in the southern areas (<40°N) over a wide range of longitudes (western and central Pacific). In contrast, the PL fishery operated in the northern areas (25–45°N) of the western Pacific. They share fishery grounds, although their hotspots differed ( Figure 2 ). Hotspots of LL were mainly found in subtropical to temperate areas (25–37°N) in the central Pacific (170–150°W) off Japan at approximately 30°N and those of PL were identified in northern areas (>42°N) in the western Pacific.

All four recaptured tagged ALB displayed an eastward movement after release ( Figure 3A ; Table 1 ). One of these (Tag ID. 2330) covered a wide range of longitudes with eastward movements from off-Japan areas to the sea area around 40°N and 155°E before returning to Japan, covering more than 25,000 km in 313 days. The remaining three individuals (Tag ID. 2337, A1132, A1127) were recaptured relatively shortly (44–78 days at liberty) but still showed migratory paths similar to those found in Tag ID. 2330.

Monthly distributions of tagged ALB showed that individuals were consistently staying around off Japan in spring (February to April) and departing the areas in May ( Figure 3B ). The ALB experienced a well-mixed layer in this area and showed strong diurnal movements distributing at 300–400 m in daytime and at 0–100 m at night ( Figure 1A ). From the one that spent 313 days before its recapture, it was shown that in summer (July to September), the ALB moved further eastward with a slight latitudinal movement to the north in certain areas (155°E and 170°E). During the eastward migration, the horizontal distribution of ALB was in the northernmost region (42°N), where a strong thermocline developed ( Figures 3B , 1B ). The ALB was mainly distributed in the surface layer (0–30 m), with occasional dives below the thermocline ( Figure 1B ). From fall to winter (October to January), the ALB returned to Japan through a route in the southern areas from the path through their early migration towards the east ( Figure 3 ). During the westward migration, the ALB was distributed at 0–100 m with dives exceeding 300 m ( Figure 1C ).

We identified T_min using sea surface temperature data from historical fishery records ( Figure 4A ) and daily water temperature through tagged ALB migration, including the northern limit of their surface habitat ( Figure 4B ). The lowest water temperature between the two datasets was extracted from tagging data (fishery data: 16.0°C and tagging data: 13.0°C). This comparison of the SST with daily water temperature from tagging data revealed that SST from logbook data is not sufficient to capture the T_min even using the large data set, as ALB distributes below the surface ( Figure 1 ). Thus, the T_min was set to be 13°C from tagging data. For the fishery dataset in LL (N=529,394) and PL (N=19,796), SST showed similar peaks (LL: 22.0°C, PL: 21.4°C) for both fisheries ( Figure 4A ), while SST in LL recorded warmer temperatures than those in PL reflecting its operation in southern areas ( Figure 2A ). Maximum SST for PL and LL are 28.6°C and 32.6°C, respectively. For the tagging data, the highest was 20.9°C with a high frequency observed at 18.6°C (N=494) mainly from the individuals staying in the off Japan areas ( Figures 3A , 4 ), whereas longtail distribution for the lower temperature came from the individual (Tag ID. 2330) migrating to northernmost areas.

One of the important steps in clustering is to determine the number of clusters objectively. The BIC was compared among candidates with grouping numbers 1 to 10. The models with one cluster have a BIC value of 26,503. The BIC did not change with two clusters, but a drastic decrease in BIC (16,031) was observed when three clusters were chosen, representing the lowest value among all the candidates. From clusters 4 to 10, the BIC values ranged from 20,371 to 38,405, with the second-lowest value occurring in cluster 8. Therefore, three clusters were used as the best grouping numbers in this analysis.

Analysis of the three clustered ocean habitats (HABs) revealed distinct horizontal separations and boundaries among them ( Figure 5 ). These HABs were separated by latitude ( Figure 5A ). HAB1 (N = 185) was characterized by a thick surface mixed layer with a sea surface temperature of 19.2 ± 0.5°C (average ± standard deviation) and occupied areas off Japan from 25 to 35°N in January to May ( Figures 5A, B ), which is known as subtropical mode water (Yasuda, 2003; Oka et al., 2023). It is not surprising that the HAB1 had the warmest among HABs, as it is located in southern areas. Notably, the average temperature at a depth of 400 m corresponded to temperature of 16.3 ± 0.8°C, warmer than the T_min ( Figure 5B ). Specifically, D_Tmin in HAB1 was below 400 m. Focusing on the vertical distribution of ALB, it was broadly distributed in the mixed layer from the surface to 400 m in HAB1, and 90% of its cumulative distribution was above 239 m ( Figure 5A ). The broad distribution was also confirmed from the individual behavior ( Figure 1A ) with changing distribution depth on a daily scale, as shown in Figure 1A (e.g., the approximate daily depths on May 5, 8, 9, and 10 were 400, 300, 200, and 100 m, respectively). HABs 2 (N= 102) and 3 (N= 101), in contrast to HAB1, showed moderate and strong thermoclines, respectively. Most HAB2 was located east of 150°E along a latitude of 35°N (between HABs 1 and 3). The others were patchily found west of 150°E ( Figure 5A ). This is mainly due to moderate thermoclines of boundary areas being classified into HAB2, as the large variance of thermal profiles indicates (the temperature at 0 and 100 m is 17.9 ± 1.5°C and 14.5 ± 2.9°C, respectively; Figure 5C ). D_Tmin , which corresponds to the depth at which the average temperature matches T_min , was 140 m. The vertical distribution in HAB2 was more concentrated at depths above 100 m, and cumulative distribution counts up to 90% were 109 m deep ( Figures 1C and 5C ). HAB3, characterized by the strongest thermocline among the HABs, occurred in the northernmost region ( Figure 5A ). Thermal gradient from the surface (16.3 ± 1.4°C) to a depth of 40 m (15.5 ± 1.8°C) was weak. Still, the gradient got stronger from the depth of 40 m to 100 m (7.9 ± 1.7°C) ( Figure 5D ). Ninety percent of the cumulative distribution (33 m) in HAB3 was above the thermocline of 40 m ( Figures 1B and 5D ), with D_Tmin occurring at 53 m.

Given the small number of tagged individuals (four) and the fragmentary nature of the dataset (2002 and 2004), it is difficult to assume that these migration routes are representative of the horizontal migration of ALB. However, discussing the universal physiological and ecological responses of species to water temperature relies on the data quality rather than the quantity of data successfully obtained from wild individuals (Holland et al., 1992; Block et al., 1997; Nakamura et al., 2020). Data quality in this study was secured, as the observed vertical distributions had a general trend of diurnal vertical migration commonly found in ALB (Childers et al., 2011; Cosgrove et al., 2014). This suggests that the effect of tag attachment on behavior was negligible. Considering the nature of our data (poor amount but secured quality), we decided not to discuss horizontal behaviors as our limitation. Instead, we focused on the universal response of species to unfavorable thermal environments.

In addition, the high-latitude areas of HAB2 and HAB3 were heavily dependent on information obtained from one individual (Tag No. 2330) for which long-term data were provided. The nature of the data obtained from one individual in a certain year should be carefully noted, although the vertical distribution obtained from this individual showed a similar trend to that reported in the Southern and Eastern Pacific Oceans (Childers et al., 2011; Williams et al., 2015), where the distribution becomes shallower at higher latitudes. It is theoretically possible that the behavior observed in this individual reflects the ecology of ALB as a species rather than a characteristic unique to the individual. Promising future research direction to overcome the limitation of tagging studies (i.e., data quantity issue) involves using otolith trace element analysis (Macdonald et al., 2013) and Iso-logging (Matsubayashi et al., 2022), to estimate the track of movement simply by collecting substantial biological tissues. Unlike tagging research, these methods are not affected by the recapture rates and therefore could be a useful approach (Macdonald et al., 2013; Matsubayashi et al., 2022) to evaluate the migration in a more quantitative manner.

Our result of minimum daily average temperature (13°C) would be regarded as the T_min of juvenile ALB. Thermal tolerance is the temperature at which a predefined ratio of individuals survives after being exposed for a certain period in a controlled laboratory environment (Sunday et al., 2011). The same definition cannot be applied because tagged individuals always survive in the wild. However, the extent of wild distributions within their thermal tolerance (Sunday et al., 2012) and the good agreement in thermal tolerance between laboratory and field studies for marine ectotherms (Payne et al., 2016) suggest that the field-based assumed minimum temperature in the northern distributional limit can be close to T_min . The T_min of skipjack tuna estimated in a laboratory study (Dizon and Yuen, 1978) was consistent with the lower average daily temperature obtained from tagged individuals in the northern distributional limit (Barkley et al., 1978; Kiyofuji et al., 2019), supporting the validity of the extraction of T_min from the daily average temperature in the distributional limit. Although other tagging studies do not apply the same definition of daily mean temperature as in our study, it is interesting to compare the minimum mean temperatures. This highlights the importance of the effect of size or age on the estimate of T_min ; 14.9 ± 1.2°C (65–90 cm FL, Childers et al., 2011);15.4 ± 0.8°C (78 cm FL, Cosgrove et al., 2014) in juvenile, and 11.3°C in adult ALB (age>5 years, Boyce et al., 2008). The thermal tolerance of ALB may be associated with growth, and our data obtained in the northern limit of the habitat led to relatively low T_min compared to other studies. We therefore regarded the 13°C as the T_min for juvenile ALB in this study ( Figure 6 ).

The shallower vertical distribution at higher latitudes could be a behavioral response to D_Tmin, which shallows at high latitudes ( Figure 6 ). Prolonged exposure to an environment below T_min leads to a decrease in cardiac output due to bradycardia (Landeira-Fernandez et al., 2004; Shiels et al., 2015), making it difficult for ALBs to maintain cardiac function. Spending most of the daily time at the surface in high latitudes, such as HAB3, would result in avoiding D_Tmin . However, deep D_Tmin in HAB1 enabled ALB to access a broad vertical range within the water column ( Figure 6 ). As tuna descend to or below the depths of D_Tmin, especially at high latitudes, a quick ascent is required to recover the cardiac function and temperature of other parts of the body (Shiels et al., 2015). This behavior is also known as behavioral thermoregulation and is found in several species (e.g, Holland et al., 1992; Kitagawa et al., 2001, 2004). Shortening the dive duration reduces heat loss from the body (Kitagawa et al., 2004). Access beyond the D_Tmin in ALB has been observed in other areas (Childers et al., 2011; Muhling et al., 2022), but body temperature drops are relatively rare owing to the short dive duration (Muhling et al., 2022). The data obtained in this study also support the behavioral response to deep cold water. In addition, the change in vertical habitat with latitude ( Figure 6 ) was also supported by the gut ALB contents; pelagic species of the Japanese anchovy (Engraulis japonicus) and midwater species of crustaceans and amphipods were dominant in HAB1 and HAB3, respectively (Kawasaki and Asano, 1962; Watanabe et al., 2004).

Latitudinal differences in the vulnerability of ALB to certain fisheries would be introduced by overlapping depths between ALB vertical habitats and fishery-specific gear setting depth ( Figure 6 ). Japanese LL fisheries tend to set their hooks deep (25–300 m or more) (Ward and Hindmarsh, 2007). In the equatorial central Pacific, the ALB catch rate increases with depth within 80–280 m (Nakano et al., 1997). The vertical distribution of ALB in HAB1, ranging from 200 to 400 m, is consistent with the hook depths of the Japanese LL, suggesting that the fishing hotspots found in the southern area coincide with ALB deeper habitat, thus requiring deeper setting of LL hooks. In HAB3 however, ALBs are more vulnerable to shallow fishing gear because their thermal habitat is contracted towards the surface ( Figure 6 ). More efficient surface capturability in this region is also well known to fishermen who constantly observe the sea surface with long-range binoculars, looking for schools of tuna (Okamoto et al., 2013; Matsubara et al., 2022). Furthermore, the short dive duration of ALB in this area may further improve the fishing efficiency of the surface fishery, as there are fewer chances of losing target fish schools owing to ALB dives. Therefore, PL hotspots appeared in northern areas near the northern habitat limit ( Figure 6 ), where strong thermoclines develop.

The strong thermoclines in HAB3, working as a thermal barrier for ALB, would also retain other tuna at the surface regarding their vertical distribution, resulting in the HAB3 area being one of the best tuna fishing grounds. Juvenile skipjack, as one of the warm tuna group members (Bernal et al., 2017), having a T_min of 18°C tends to stay at the surface in this area (Kiyofuji et al., 2019), and a similar surfacing pattern is also found in bluefin tuna of the cold tuna group which has the T_min of 14°C (Kitagawa et al., 2001; Fujioka et al., 2018). For both warm and cold tuna groups (Bernal et al., 2017), the thermocline that developed in the area around 40°N was strong enough to restrict their vertical distribution on the surface ( Figure 6 ). High productivity in the Kuroshio–Oyashio extension area (Yasuda, 2003; Sassa et al., 2007) has the potential to meet the high energy demand for tuna (Kitagawa et al., 2007; Aoki et al., 2017, 2020; Muhling et al., 2022) and aggregate them in this area (Aoki et al., 2017) though however, the vertical utilization of the habitat in the area is limited. Higher productivity on these fishery grounds has been interpreted as the effect of the mixing of the warm Kuroshio and cold Oyashio currents (Yasuda, 2003; Zainuddin et al., 2004). In addition, our results underscore the importance of species-specific physiological thermal tolerance in the formation of rich fishery grounds.

This study introduces a novel aspect to understanding the relationship between the vertical distribution of juvenile ALB and their vulnerability to fisheries by examining the response of this species to its T_min , in the wild. It elucidates how the vulnerability of ALB to fisheries and catch variability are closely related to their physiological tolerance, vertical habitats, and D_Tmin ( Figure 6 ). Integrating the responses of ALB to thermal profiles with fishery information such as CPUE (Williams et al., 2015; Mondal et al., 2022) will enhance future stock assessments, especially in boundary areas characterized by D_Tmin within the species’ vertical habitat. Furthermore, analyzing these biological and behavioral responses will serve as a crucial reference for projecting the effects of climate change scenarios (Sunday et al., 2012) to assess its impact on fisheries and stocks (Lehodey et al., 2015), thereby reducing uncertainty in future stock management to maintain sustainable fisheries.