An individual-based model of North Pacific albacore tuna seasonal migratory behaviour and climate sensitivity

Abstract

Juvenile North Pacific albacore tuna (Thunnus alalunga) undertake annual long distance migrations between offshore waters and the California Current Large Marine Ecosystem (CCLME), yet the drivers of the timing of these movements remain unclear. Highly migratory marine predators like albacore often use environmental cues to track seasonal resources and optimize foraging. Mixed layer depth (MLD), defined as the well-mixed surface layer of the ocean, has previously been associated with important albacore physiological and behavioral patterns. Using electronic tagging data and an individual-based model (IBM) we show MLD has a pivotal role in influencing albacore migration timing and depth preferences. Albacore actively expand their vertical habitat in correspondence with wintertime MLD deepening and appear to utilize a 30m MLD threshold to initiate preemptive movements to reach seasonally and spatially explicit foraging resources. Model simulations using MLD-based rules and an ocean sea surface temperature (SST) constraint successfully capture the seasonality of movements and distribution of albacore. Climate projections suggest that by 2070–2099, SST warming will shift albacore distributions poleward and MLD shoaling will prolong their coastal residence, potentially increasing albacore concentrations in the Northern CCLME. These findings highlight the relevance of subsurface ocean conditions to the movement of highly migratory species and demonstrate the utility of IBMs in the study of complex migratory behaviors.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-46968-y.

Introduction

Highly migratory marine predators undertake long-distance movements that are both critical to their survival and threatened by increasing environmental pressures. Migration is an evolved adaptive behavior predominantly concerned with maximizing feeding and reproductive successes^1,2. Hundreds of species of birds, fish, mammals and reptiles regularly travel across oceans to make use of favorable oceanic conditions to breed and raise their young as well as feed in productive ecosystems^3–5. The quality of these resources often varies in time and can be spatially distant, requiring animals to perform well-timed migrations to follow seasonal resource waves or to connect breeding and foraging sites^6,7. Complex circumpolar or trans-oceanic migrations among various taxa suggest that a wide range of mechanisms such as memory, perception of environmental cues, and geomagnetic navigation are used to successfully perform these movements^3,5. Migrations are inherently energetically risky, especially if the journey requires passing through regions with few resources or if environmental variability causes a mismatch in timing between arrival and favorability in conditions^1,8. Humans have increased pressures on many migratory species via overexploitation, fisheries bycatch, habitat disruption and pollution, and anthropogenic climate change^8–10. Migratory animals are some of the most at risk from rapid environmental changes because of their reliance on geographically separated resources and habitats^11,12, making it critical to understand the dominant processes and factors underpinning the decisions driving migrations of these species.

Albacore tuna (Thunnus alalunga: albacore hereafter) are known for their extensive migrations that are influenced by changes in oceanographic conditions and prey availability. Albacore are a globally distributed temperate and tropical predator that are both ecologically and commercially important¹³. Their distinct thermal preferences dictate much of their population-scale distribution^14–17. Every year in the North Pacific, some juvenile albacore migrate between coastal foraging habitat in the California Current Large Marine Ecosystem (CCLME) and offshore habitat spanning as far westward as Japan^14,18. Favorable temperatures and high prey abundance in the summer and fall across the CCLME attract albacore and other species to the North American coast³. Once food becomes scarcer and environmental conditions change, many albacore migrate west, incurring significant energetic costs in exchange for profitable foraging opportunities in productive offshore areas^14,18,19. Albacore’s highly adaptive and variable foraging behavior allows them to efficiently exploit these diverse ecosystems and balance the risks and rewards of migration²⁰. While albacore migrations are primarily focused on accessing food and favorable conditions, the specific drivers of migration timing are poorly understood.

One potential environmental factor driving albacore cross-Pacific migrations that has yet to be fully explored is mixed layer depth. Mixed layer depth (MLD) refers to the thickness of the upper layer of the ocean where temperature and salinity are relatively uniform and the water is well-oxygenated due to wind and wave-driven mixing²¹. MLD is known to be an important environmental variable influencing albacore vertical depth preferences because of the species’ physiological limits^14,16,19,22. Albacore generally use the warmer mixed layer as a thermal refuge at night, conducting deeper dives to forage during the daytime, with foraging time below MLD dependent on the thermal profile of the water column^14,16,23. The seasonal cycle in MLD across the Pacific has also been associated with different albacore foraging behaviors. Albacore feed on shallower prey layers in the CCLME during the summer and fall, which aligns with shallow MLDs^14,20,24. Many juvenile albacore forage in the offshore Pacific during wintertime, when the mixed layer is deeper and they can access organisms associated with shallow and deep scattering layers^25,26. This suggests that deeper MLDs may expand the vertical habitat available to albacore, allowing the fish to feed on additional prey and​​ offset the large energetic demands from their lengthy migrations^18,19,23.

Given the influence of oceanographic conditions on albacore behavior, climate change may further impact their distribution and migrations through the rest of the 21st century. Oceans are projected to see significant warming²⁷ and increased stratification with an associated shoaling of the mixed layer²⁸, which have large ecosystem-scale implications regarding biological limits, primary productivity and food web dynamics²⁹. Historical analyses of albacore catch data have already shown incrementally earlier migrations in the Northeast Atlantic over the last 60 years, in part due to changes in hydro-meteorological conditions and rising ocean temperatures^30,31. Studies employing various different models have predicted poleward shifts in the Pacific^32,33 and potential decreases in biomass and body size by 2050³⁴. Conversely, movement of favorable thermal habitat towards the poles may allow albacore to access highly productive sub-arctic ecosystems³⁵. Understanding how spatial distributions of commercially important, highly migratory marine species like albacore might be affected by future environmental changes is critical for fisheries management and ecosystem conservation.

A large part of the ecological modeling literature on highly migratory species such as albacore makes use of species distribution models (SDMs), which find statistical relationships between animals and their environment to determine spatial extent of suitable habitat^36,37. While these models have been successfully used in spatial ecology and conservation, they may experience deterioration in performance and large uncertainties when extrapolated to novel environmental conditions^38–42. Individual-based models (IBMs) instead offer a more mechanistic approach to study the behavior and decision-making processes of marine predators. These models use deterministic or stochastic rules based on pre-selected drivers to simulate the movements of individual animals, allowing researchers to investigate the underlying processes that drive behaviors. Ensembles of many individuals can then be used to analyze emergent population dynamics in spatially and temporally dynamic environmental settings. The flexibility of IBMs has been demonstrated in studies that reproduced interannual variability of sea lion foraging success in the CCLME⁴³, temperature and prey-driven migration patterns of blue whales⁴⁴, and long-term distribution of skipjack tuna in the South Pacific⁴⁵.

This study investigates how MLD influences vertical habitat preferences and migration patterns of North Pacific juvenile albacore by using data from albacore fitted with archival tags to build an IBM. The complexity of the IBM is progressively increased by incorporating temperature and MLD rules based on analysis of the tag data. A climate change analysis is performed by simulating albacore movements under projected future conditions. This research aims to expand our understanding of ecological drivers involved in the migratory processes of albacore and provide insights about possible spatial and phenological shifts caused by changes in ocean conditions.

Results

Distribution and seasonal migration patterns

Location data from the 12 tagged juvenile albacore tuna that undertook long-distance migrations reveals a dominant seasonal pattern in the movements, which can be divided into four phases. In January through March (winter phase) albacore were found offshore in the North Pacific Transition Zone (NPTZ) and Subtropical Gyre (NPSG; Fig. 1A). From April to June (spring phase), albacore departed from their offshore foraging habitat and migrated east towards the CCLME, with certain individuals traversing more than 5000 km and sustaining the fastest average speeds of any three month period (Fig. 1B; example tracks found in Supplementary Fig. 1). Between July and September (summer phase), albacore were observed almost exclusively within the CCLME where they foraged in their nearshore coastal habitat (Fig. 1C). Finally, in October through December (fall phase) albacore migrated west, with all but three individuals leaving the coast by the end of the year (Fig. 1D).

Fig. 1.

A-D, Locations of 12 tagged juvenile albacore (pink) during the four phases of their North Pacific migrations. The tracks are overlaid on the monthly climatology of mixed layer depth and bounded by temperature preference range (blue envelope), both of which are averaged across the three months of each phase. The temperature range is constant throughout the year and spans from ~ 13 °C to ~ 20 °C. The arrows indicate the magnitude direction of albacore movement, westward (blue) and eastward (red). E-F, Monthly mean SST and MLD experienced by tagged albacore at their locations, with standard deviation error bars. SST and MLD are sampled from the GLORYS12V1 product at the time and location of each datapoint with spatial resolution of 1/12^o and 1^o, respectively. G-H, Monthly mean longitude and zonal velocity (East-West) recorded by tags.

This pattern is primarily observed in 8 of the 12 albacore, all of which were captured and released in the Northern CCLME. The remaining tracks showed greater variability in departure and return time from their coastal residence, with 3 individuals originating from the Southern CCLME and one migrating into the Southern CCLME from a Northern origin (example tracks found in Supplementary Fig. 2). These fish migrated offshore for shorter periods of time (94 versus 234 days), did not travel as far (1872 versus 4860 km) and utilized offshore waters in the NPSG closer to the coast compared to the Northern fish. Regardless of their migration patterns, albacore showed a strong sea surface temperature (SST) preference, with 75% of data points found in waters with SST between 14 and 18.5 °C and > 95% between 13 and 20 °C. Their latitudinal habitat range was largely dictated by this thermal preference, resulting in North-South shifts of the thermally favorable portion of the North Pacific throughout the year (Fig. 1A-D).

Mixed layer shoaling or deepening to an average depth of 30 m in the region where the tagged albacore were located coincided with the commencement of their large-scale eastward and westward movements, respectively. The majority of albacore migrated towards their coastal habitat when the MLD shoaled to this threshold depth, and their zonal velocity increased as the MLD became progressively shallower in the following 3 months (Fig. 2). After their summertime residence in the CCLME, their offshore migration aligned with the mixed layer deepening to 30 m and their westward movements continued for several months as the MLD continued to deepen.

Fig. 2.

A, Zonal velocity of tagged albacore highlights the four distinct phases of the trans-Pacific migration. B, Monthly climatologies for the MLD in the inshore and offshore North Pacific (blue) calculated by averaging the GLORYS12V1 MLD product across 2004–2014. The inshore region is defined as the area where the mean annual SeaWiFS is > 1.5 mg m^{− 3} in the CCLME. CHL-a Monthly climatology for productivity in the CCLME (green) is represented by the time series of the first EOF of SeaWiFS chlorophyll concentration⁴⁶. Vertical lines show the commencement of eastward migration (red) and westward migration (blue) aligned with the MLD shoaling or deepening to the 30 m MLD threshold, respectively.

Vertical habitat and MLD

High-resolution depth data recorded at one-minute intervals reveal significant seasonal shifts in albacore vertical habitat that followed changes in MLD and geographical location. Albacore consistently exhibited diurnal diving, with average daytime depths deeper than average nighttime depths (58.6 versus 21.5 m; Fig. 3A). While this behavior was observed year-round, albacore occupied deeper layers in winter and spring compared to the summer and fall. Daytime albacore depths closely tracked the deepening of the MLD as they left the CCLME, peaking in February when MLD was deepest, and then decreasing throughout the spring as the MLD shoaled and the fish migrated back to their coastal habitat (Fig. 3A). A similar pattern is observed in nighttime depths, as their nighttime vertical habitat expanded and contracted according to the MLD. Albacore remained relatively shallow whilst inside the CCLME, with no observed difference between albacore found in the Northern or Southern portions of the ecosystem (Fig. 3C). A visible northeast to southwest gradient in albacore depths reflects the dominant migration pattern in the same direction that is followed by almost all of the Northern CCLME individuals (Fig. 3C). Daily average depths are greatest in the NPSG, which is largely inhabited by albacore in their offshore residence when the mixed layer is deepest (Fig. 3C-D).

Fig. 3.

A, Average daytime depths (red) and nighttime depths (blue) for each day of the year and average monthly MLD experienced (black) across all 12 tagged albacore. Months when albacore inhabit the CCLME are shown in green. B, Albacore dive depths are calculated by averaging the top 10th percentile daytime and nighttime depths of each individual. Dive depths largely correspond to the depth bounds of albacore vertical habitat at night and day. C-D, Approximate daily locations of albacore tracks colored by their average daytime depth and average MLD, respectively.

Daytime albacore dive depths, which indicate foraging depths, show a marked distinction in foraging strategies between periods of offshore and coastal residence. During their CCLME residence from July to October, the fish foraged at shallow depths and spent 95% of the day within the top 100 m (Fig. 3B). Once out of their coastal habitat, albacore switched to deep-diving foraging, doubling their average dive depths and foraging below 100 m for almost 30% of the day (Fig. 3B). While on average albacore depths gradually decreased as the MLD shoaled during their coastward migration between April and June (Fig. 3A), albacore continued exhibiting deep-dive foraging during this period (depths > 150 m; Fig. 3B). This resulted in albacore spending > 30% more time during the day and > 50% more time during the night below the MLD compared to the rest of the year. The fish only switched back to shallow foraging once they had entered the CCLME. These transitions between surface-dwelling foraging and deep-diving foraging occur as albacore move into and out of the CCLME and happen rapidly in less than a month (Fig. 3B).

Individual-based model simulations

The individual-based model (IBM) of the migratory behavior of juvenile Pacific albacore incrementally increases in complexity to investigate whether including environmental drivers better captures albacore habitat and behavior. First, the base model generates albacore tracks using standardized random velocities based on the statistics of the tagged albacore and a land buffer (see Methods). Without rules aimed at replicating environmentally-based decision making, the fish disperse throughout much of the Pacific and into regions unexplored by the tagged albacore (Fig. 4A). The range of temperatures experienced by the simulated albacore is significantly broader than the observed range (Fig. 4B) and no seasonality in longitude and zonal velocity is recorded (Fig. 4C).

Fig. 4.

IBM simulations of albacore tracks (120 tracks, n = 50760) of each model iteration. A-C, Null hypothesis model uses standardized random velocities and a land buffer. D-F, Temperature constraint model uses a stochastic temperature constraint to model albacore thermal preference. G-I, Full swim model incorporates 2 MLD-based rules along with the temperature constraint. Realistic temperatures, zonal velocities and average longitudinal position are captured.

The next model iteration includes a probabilistic temperature constraint that moves the fish away from waters deemed unfavorable based on their thermal preference range (see Methods). With this sea surface temperature constraint, the ensemble of simulated tracks begins to fit the overall spatial distribution of the tagged tuna (Fig. 4D) and maintain the tracks within a realistic temperature regime (Fig. 4E). However, the simulated albacore predominantly remain in the Eastern Pacific and exhibit no seasonality in the direction of their longitudinal movements (Fig. 4F).

Informed by the relationship between albacore zonal velocity and MLD observed in the tagged data, two rules that adjust the zonal movements of the fish are finally incorporated into the full swim model. The direction and magnitude of these adjustments are based on the 30 m MLD threshold previously discussed and whether the mixed layer is deepening or shoaling (see Methods). Implementing this final set of rules generates tracks with overall realistic spatial distribution (Fig. 4G) and similar longitudinal movement and position (Fig. 4I) with respect to the observed albacore. Distributions of the simulated fish using the full swim model capture the regions inhabited by the tagged fish during their coastal and offshore residences (Fig. 5A and C). The model is also able to recreate the timing of the start of the spring and fall migrations and the longitudinal range explored by the tagged albacore during their long distance movements (Fig. 5B and D).

Fig. 5.

IBM simulated albacore tracks using the full swim model compared to observed albacore tracks split into the four migration phases (A: spring, B: summer, C: fall, D: winter). The observed tracks are overlaid on albacore densities simulated under the reference (2003–2013) scenario. Albacore densities are calculated on a 1.5^o grid and display the number of albacore present in each grid cell.

Climate change sensitivity

To explore basin-scale changes in albacore distribution and migration patterns due to climate change, the full model is run on 2070–2099 climate projections for SST and MLD under the SSP2-4.5 concentration scenario. Ocean temperatures are projected to increase while MLD is largely projected to shoal, especially in the Western North Pacific (Supplementary Fig. 3). The warmer year-round temperatures result in a poleward shift of the thermal preference envelope for albacore across all four seasons. Under these conditions, the model predicts that albacore will access newly available habitat off the coast of Canada and in the NPTZ during their coastal residence and migration periods, respectively, and avoid offshore waters in the NPSG that are projected to reach temperatures outside of their temperature range (Fig. 6B).

Fig. 6.

A, Simulated albacore density (120 tracks, n = 50760) in the North Pacific under reference (2003–2013) conditions. B, Difference between simulated albacore density under 2070–2099 SST & MLD projections and reference scenario. Habitat gain (dark red) shows where albacore are found in the future scenario but not in the reference scenario while habitat loss (dark blue) signifies the opposite. C-F, Comparison of arrival and departure into the CCLME, time spent in the CCLME, and maximum migration distance. Time spent in the CCLME was only calculated for albacore that left and entered the coastal ecosystem at least once (88 simulated tracks) to allow for cross-scenario comparison. Migration distance was calculated as the maximum distance from the point of origin of the track. The median value from the 12 tagged albacore tracks is shown by the star. The spatial maps and barplots for the 2070–2099 SST scenario (maintaining present MLD conditions) and the 2070–2099 MLD scenario (maintaining present SST conditions) can be found in Supplementary Fig. 4.

Concurrently, shallower mixed layers are projected to affect the timing of albacore migrations, prolonging their residence in the CCLME (Fig. 6C-E). Under future conditions, the 30 m threshold MLD associated with the start of their seasonal migrations occurs earlier in the spring and later in the fall. According to the IBM, this shoaling would result in earlier coastward migrations by a median of 15.3 days (95% CI: 5–26 days) and delayed offshore migrations by a median of 20.8 days (95% CI: 12–29 days) for the simulated tracks. For simulated albacore tracks that include exit and re-entry into the CCLME, their coastal residences increase by a median of 45.75 days (95% CI: 24–63 days) relative to simulations under reference conditions (Fig. 6E). Due to the shorter period available for these simulated future albacore to migrate offshore, their maximum migration distances also decrease by more than 400 km (Fig. 6F). The combination of the prolonged coastal residence and the poleward shift results in the greatest increases in albacore densities to occur in the Northern CCLME (Fig. 6B).

Discussion

In this study we investigated hypotheses that juvenile albacore migrations in the North Pacific are influenced by SST and MLD by analyzing long-term tagged juvenile albacore tracks and utilizing an IBM. Known migratory patterns^14,18, vertical behaviors¹⁴, physiological limitations and migration phenology²³, and foraging ecology^19,20,24, when examined together, suggest that changes in MLD may act as an environmental cue for albacore migration timing between coastal and offshore habitats. The IBM developed with SST and MLD-based rules reliably captured the temporal and spatial dynamics of juvenile albacore migrations across the North Pacific. Analysis of albacore depth data reveals albacore expand and contract their vertical habitat according to seasonal changes in MLD, providing further causal support for the model. This mechanistic model allows for projections of albacore migration timing and distribution under future conditions, providing us a tool to evaluate population-scale impacts of climate change.

Although sea surface temperature is known to constrain albacore habitat, this variable only partially informed albacore migration according to our IBM. Tagged albacore actively sought out favorable ocean temperatures, moving North-South throughout the year to stay within their temperature range and exhibiting no strong seasonality in sea surface temperatures experienced. Iterations of the IBM incorporating a probabilistic SST-constraint captured these large-scale latitudinal shifts, but failed to recreate the rapid longitudinal cross-Pacific movements observed in many of the tagged individuals. Previous work proposed movements of thermal fronts and the warming of coastal waters as factors driving the coastward shift of albacore catch near the US West coast¹⁷. While these smaller-scale movements may be influenced by changes in sea surface temperatures, there are no significant longitudinal thermal gradients that could drive migrations across 1000 km of open water. In fact, Muhling et al., (2022)¹⁸ pointed out that tagged albacore migrated much further offshore than required solely based on their thermal preferences, suggesting that these movements were aimed to anticipate favorable foraging opportunities. Our results suggest that thermal limits influence large-scale latitudinal albacore distribution but cannot explain these longitudinal migrations.

The IBM was able to replicate the longitudinal distribution of albacore and the timing of their preemptive migratory movements only when seasonal changes in MLD were included in the model. To transition between nearshore and offshore waters in time to make use of optimal conditions, albacore began their migrations several months in advance of summertime peak coastal productivity and deep wintertime offshore MLDs. Our findings demonstrate that the initiation of these movements aligns closely with the shoaling and deepening of MLD across an estimated threshold of 30 m depth. Movements become more directional and travel speeds increase significantly during these periods, allowing albacore to minimize their time spent traveling between habitats. Introducing two simple MLD rules based on this threshold to the IBM successfully reproduced these longitudinal migration patterns. Tagged albacore appear to be using the seasonally predictable transitions between shallow and deep mixed layer domains as reliable signals to optimally time their departures. The ability of our model to reproduce the seasonal movements of tagged albacore demonstrates that the predictable seasonal changes in mixed layer depth in both nearshore and offshore albacore destinations²³ not only influence energetics and habitat availability for albacore^18,19, but can explain and predict the phenology of albacore migration.

IBM results are consistent with our analyses of albacore vertical habitat use, further underscoring the role of mixed layer depth (MLD) in shaping albacore distribution. As tagged albacore migrated offshore into the central North Pacific, they occupied progressively deeper depths and increasingly spent time below 100 m, aligning with the seasonal deepening of the mixed layer and a shift in albacore prey availability. In nutrient-rich coastal systems such as the CCLME, higher concentrations of epipelagic prey near the surface support albacore surface foraging on shallow prey layers^20,24. In contrast, prey in oligotrophic offshore waters tend to be located deeper, requiring albacore to adopt deep-diving foraging strategies¹⁹. The deeper wintertime MLD in these offshore regions enables albacore to remain closer to these prey layers while reducing time spent in colder waters below the mixed layer, thereby lowering energetic costs. Indeed, Arostegui et al., (2023)¹⁹ found that albacore expended up to 26% less energy when foraging in offshore habitats compared to nearshore environments, primarily due to reduced exposure to cold sub-mixed-layer temperatures. The seasonal favorability of offshore biomes is further supported by the winter foraging patterns of other deep-diving predators, highlighting the broader ecological significance of these regions in shaping migratory behavior^3,5,47,48. Interestingly, many of the fish continued exhibiting deep-diving foraging between April and June as they migrated towards the coast despite significant MLD shoaling. While this behavior represents an energetically costly strategy¹⁹, albacore recorded the highest foraging success during this period, potentially offsetting the added energetic costs¹⁸. This suggests that offshore regions in the NPTZ and NPSG offer sustained foraging opportunities until increased productivity and warmer SST return to the CCLME during the summer and early fall. Once the albacore arrived in nearshore waters, they inhabited shallower depths as the mixed layer continued to shoal.

Model limitations are observed when simulating albacore individuals originating from the Southern CCLME, suggesting regional differences may alter migration patterns. The small subset of individuals originating from southern waters (n = 3) undertook less extensive movements and exhibited greater variability in migration phenology and duration. The 30 m MLD threshold failed to match the timing of their migrations, resulting in the model simulating tracks that differed significantly from their respective tags. While not included in this study, a sizable portion of the Southern population remains resident year-round off the coast of the Baja peninsula¹⁸. Warmer winters along with less-pronounced seasonal cycles in environmental conditions for Southern albacore likely contribute to these permanent or prolonged coastal residences as well as the less precise timing in migrations²³. The Southern individuals also were larger and closer to maturity compared to their Northern counterparts, potentially driving shifts in behavior focused on allocating energy towards reproduction and spawning migrations. These factors may result in southern individuals utilizing alternative signals to initiate their movements, meaning our simple MLD-based model is capturing migration drivers specific to the Northern individuals. While previous work has confirmed these albacore are from the same genetic stock⁴⁹, significant behavioral plasticity between and within geographical populations adds further complexity. Additional research should address the regional environmental and physiological differences, exploring how differences in local prey dynamics, temperature, and MLD influence migratory decision-making and adaptability to changing ocean conditions.

According to the mechanisms utilized in the IBM, albacore migrations were predicted to shift both spatially and temporally with climate change. Assuming that these albacore retain similar thermal preferences and continue following the same MLD signals during their migrations, future albacore populations can be expected to undergo a poleward shift from rising SSTs and prolong their coastal residence due to shoaling in MLD. Our model not only confirms temperature-driven latitudinal shifts of albacore catch along the US West coast⁵⁰, but also suggests potential large-scale longitudinal changes in distribution due to altered migration phenology. Earlier arrivals into the CCLME and delayed departures could increase potential overlap with coastal troll and pole-and-line fisheries and decrease overlap with offshore fleets. Longer residence periods within the CCLME in turn decrease the time available for albacore to migrate offshore. If offshore foraging periods and travel distances are shortened, long-distance migrations to the NPTZ and NPSG may become less energetically profitable. Marine heatwaves offer valuable analogs for future climate conditions, as they are characterized by higher SSTs and shallower MLDs^51,52. While our tagging data largely pre-date a series of marine heatwaves that impacted the CCLME starting in 2013, future work on albacore migrations during these anomalous events could provide further insights into the implications of climate-driven changes on migrations patterns.

Individual-based models (IBMs) offer a powerful framework for identifying future risks to migratory populations, particularly under climate change scenarios that may disrupt the alignment between migratory behavior and favorable environmental conditions. Unlike purely statistical models, mechanistic approaches such as IBMs explicitly incorporate the processes underlying movement and foraging behavior, allowing researchers to explore how population vulnerability emerges when different habitats change at different rates or in different ways^11,53,54. For albacore, shifts in the timing or location of peak prey availability relative to established migratory cues could reduce foraging efficiency, with potential consequences for growth and reproductive success. Climate projections already suggest declining suitability of subtropical foraging habitats and increasing suitability in sub-Arctic regions³⁵, highlighting the need for predictive tools that can anticipate such habitat transitions. Although albacore and other highly migratory species exhibit behavioral flexibility, the pace of climate change may exceed their capacity to adapt, especially if migration relies on environmental cues that no longer signal optimal conditions. Mechanistic models like IBMs are especially well-suited to investigate these risks, as they simulate individual-level responses to environmental variability and can scale those responses up to forecast population-level outcomes. As shown by Siren et al., (2025)⁴⁰, mechanistic models can outperform correlative approaches when applied to novel environmental conditions, but only when they are accurately tailored to the ecological system in question. Even small model misspecifications can lead to large errors in prediction. This highlights the importance of combining robust ecological understanding with careful model design. When developed thoughtfully, IBMs provide a critical tool for evaluating the resilience of migratory species under future climate conditions and for identifying potential mismatches that may otherwise go undetected.

Data & methods

1. Albacore tuna archival tag data

This study analyzed albacore tracks previously collected by Albacore Archival Tagging Program: a collaborative effort between the NOAA Southwest Fisheries Science Center and the American Fishermen’s Research Foundation. All tagging procedures and animal handling, as described in Childers et al., (2011)¹⁴, Snyder (2016)²³, and Muhling et al., (2022)¹⁸, were conducted in accordance with relevant guidelines and regulations of the California Department of Fish and Wildlife (permit number SC-13886). This study did not involve any new experiments on live vertebrates. Tagging procedures and initial data analyses are briefly summarized here.

Archival tags were fitted to juvenile albacore caught in the California Current System in summer and fall between 2003 and 2011. Three types of archival tags (Lotek LTD2310 and LAT2810, and Wildlife Computers MK9) were used. The tags recorded pressure, light, and peritoneal and ambient water temperatures every minute. Daily light-based geolocations were estimated using manufacturer-provided software. A sea surface temperature (SST) inclusive unscented Kalman filter (ukfsst package version 0.3^55,56), was then fitted to all tracks using the NOAA Optimum Interpolation SST V2 product⁵⁷ in R 3.6.3 (R Core Team, 2020). A bathymetric correction constraining estimated locations based on daily fish depth was then applied using the analyzepsat package⁵⁸. The first 30 days of data after release were not included in analyses to remove potential effects of tagging on behavior. Post-processed geolocation data have an estimated error in the order of 1–2°, with the longitudinal component more certain due to constraints from dawn-dusk^14,55,59.

Tagged albacore showed a variety of movement patterns, ranging from highly migratory to largely resident in the California Current System¹⁸. As this study focused on migration patterns, 12 fish that undertook long-distance migrations were included (Table 1).

Table 1.

Summary of 12 tagged juvenile albacore included in this study.

Tag	Manufacturer	Model	Release Date	Recapture Date	Release length (cm)	Days at Large
1987	Lotek	LTD2310	11/8/2003	8/21/2004	88	287
2082	Lotek	LTD2310	11/10/2003	8/20/2004	81.5	284
2393	Lotek	LTD2310	6/30/2004	5/27/2006	66	696
2381	Lotek	LTD2310	7/1/2004	9/1/2005	65	427
2398	Lotek	LTD2310	7/1/2004	9/19/2005	64	445
2942	Lotek	LTD2310	8/12/2004	7/5/2005	82.5	327
1045	Lotek	LTD2310	8/6/2006	6/16/2008	80	680
1464	Lotek	LTD2310	8/6/2006	8/21/2007	75	380
394	Lotek	LAT-2810	8/3/2011	8/15/2012	68	378
1,090,251	Wildlife Computers	Mk9	8/3/2011	8/24/2013	65	752
396	Lotek	LAT-2810	8/3/2011	5/16/2013	63.5	652
1,090,269	Wildlife Computers	Mk9	8/4/2011	6/26/2013	64.5	692

2. Environmental reanalysis datasets

Environmental data for sea surface temperature (SST) and mixed layer depth (MLD) were obtained from the GLORYS (Global Ocean Reanalysis and Simulation) 12V1 product, provided by the Copernicus Marine Environment Monitoring Service (CMEMS)⁶⁰. Model reanalysis data was determined more suitable for the purposes of this project over tag-derived data, particularly for MLD. Inconsistent sampling of the mixed layer by tagged albacore precluded estimation of MLD from tag data for all days. In contrast, the basin-scale spatial and temporal coverage provided by GLORYS enables consistent characterization of MLD across the Pacific and supports the development and testing of hypotheses regarding environmental drivers of albacore migration and their responses to oceanographic variability. An additional motivation for using GLORYS is that is provides a spatially and temporally continuous MLD field across the basin, yielding a consistent dataset for hypothesis development and evaluation. Restricting.

analyses to MLD estimates derived from tuna tags would introduce inconsistencies relative to GLORYS-based MLD, as tag-derived estimates would rely on a temperature-gradient definition alone. Such estimates are sensitive to methodological choices and would require calibration to ensure comparability with GLORYS MLD fields.

We used monthly mean fields at a horizontal resolution of 1/12° (approximately 8 km at mid-latitudes), covering the period from 31 December 1992 to 24 June 2024. To align with the temporal coverage of the tagged albacore tuna in our study, we utilized data from 19 January 2003 to 16 December 2013 and computed seasonal monthly climatologies. SST was directly extracted from the GLORYS dataset while a Shapiro filter was applied to smooth the MLD data, achieving an effective resolution of 1° (approximately 100 km). This smoothing process reduces small-scale variability, emphasizing larger-scale oceanographic features that are more relevant to albacore tuna migration patterns. In the GLORYS model, the MLD is defined as the depth at which the potential density exceeds the surface value by 0.01 kg m⁻³, following the criterion established by de Boyer Montégut et al., (2004)²¹. This definition accounts for both temperature and salinity influences on water density, providing a more accurate representation of ocean stratification.

3. Individual-based model development

To simulate the movement and migration patterns of albacore tuna, we developed an Individual-Based Model (IBM) to track the longitude and latitude positions of individual albacore over time. This approach allows for the modeling of spatiotemporal dynamics of tuna as they navigate through their marine environment, driven by various biotic and abiotic factors. The model generates tracks in 2D space across the North Pacific using a progressively complex rule-based system to produce zonal and meridional velocities, beginning with a simple random velocity model and building up to more ecologically informed movement rules. This staged approach allows us to gradually refine the model to better capture albacore tuna behavior.

These movement rules are based on analysis of geolocation data with relatively large error due to the nature of archival tagging. With this in mind, we focused on long-term movements over distances of magnitude higher than expected error and used velocities between daily-averaged geolocation datapoints that are calculated from tens to hundreds of datapoints each day. This temporal smoothing of the data reduces the impact from uncertainty that might affect analysis of highly variable sub-daily scale movements and behaviors. Additionally, the migration patterns are mostly longitudinal with individuals traveling at largely constant speeds and directions see Supplementary Figs. 1 and 2), making analysis and simulation of longitudinal movements less prone to error.

3.1 Random velocity rule

To ensure simulated velocities are within a biologically plausible range and capture swimming behaviors exhibited by albacore, we analyzed the tagged albacore velocity data and calculated daily mean and standard deviation for the zonal (u) and meridional (v) velocity components. Simulated daily velocities are randomly generated to have a probability density function (PDF) matching that of the tagged daily velocity data.

Furthermore, calculating the auto-correlation function of the velocity components showed a significant level of memory in the fish’s movement (correlation = 0.61), meaning that the direction and speed at any given time are influenced by past behavior. To capture this memory effect, we incorporated an autoregressive model of order 1 (AR-1) into the random generation of velocities. The AR-1 model maintains the correct variability in swimming speeds and directions based on statistics calculated from the tag data while also introducing realistic randomness into the movement model. The equations to generate the velocity components are:

Where regression coefficients represents the memory effect, and represent the random noise introduced at each time step, and and ​ appropriately scale the random noise to the PDF of the tagged albacore velocities. The AR-1 approach allows the generated velocities to retain the same statistical properties as the observed data, including both the PDF and the memory structure of the velocity components. This serves as a baseline for movement, with no external environmental constraints influencing fish behavior to assess general dispersal patterns.

3.2 Temperature-constrained velocity rule

In the second stage, the model integrates constraints based on upper ocean temperature approximated by SST, which is known to be an important factor influencing albacore distribution^15,17,61,62. We used temperature data from electronic tagging studies to inform the range of temperatures that the fish are likely to prefer. A custom binomial probability curve is used to assign the likelihood of albacore finding specific sea surface temperatures favorable to ensure the range of temperatures experienced by simulated albacore replicate those recorded by the tagged albacore. For example, regions with SSTs within ± 1 standard deviations of the mean were considered 85% favorable, meaning 85% of the time albacore would remain in the location and 15% they would be redirected towards more favorable temperatures using the local SST gradient. This probabilistic approach allows for simulated albacore to occasionally venture into regions with SSTs at or beyond their thermal preference range, but for the most part maintain tracks within favorable conditions that reflect the strong thermal preference that drives albacore habitat.

3.3 Mixed layer depth rules

In the final stage of the IBM, we incorporated mixed layer depth into the rule set. The goal was to create a simple and general method for simulated albacore to initiate and progress through their migration without being based on the time of year or geographical location. Informed by the observed patterns in zonal velocities, longitudinal positions and depths of tagged albacore that matched seasonal changes in MLD in the North Pacific (Figs. 1, 2 and 3), two rules were developed to drive east-west albacore movements:

1. If the MLD is shoaling and is shallower than 30 m, albacore are nudged east towards the productive regions of the CCLME.

2. If MLD is deepening and is deeper than 30 m and shallower than 70 m, albacore are nudged west towards offshore habitat.

The rules use a combination of a threshold MLD and the temporal gradient (shoaling vs. deepening) of the mixed layer to apply the nudges at the appropriate time. A MLD value of 30 m is observed to coincide with the start of albacore migrations to and from the CCLME, serving as a physical trigger that signals the transition into migratory phases. The MLD shoaling to depths shallower than 30 m aligns with albacore eastward migration, while the MLD deepening to depth greater than 30 m aligns with their westward migration. The temporal gradient of the MLD is calculated using the difference between the weekly-averaged MLD from the contemporaneous week and the week prior. Daily temporal gradients were not used to inform the model whether the albacore were experiencing a shoaling or deepening MLD as they proved to be sensitive to small changes in MLD due to the random motion of the simulated albacore, causing nudges to be applied at undesired occasions.

When either set of conditions are met, a fish’s random velocity vector is adjusted via a “nudge” towards the desired direction. The process of applying a “nudge” adds an additional velocity vector of a certain angle and magnitude, which are then renormalized to maintain the same magnitude speed of the original velocity vector. Depending on the magnitude of the nudge, the fish’s new velocity will be more or less affected by the adjustment. The westward (fall) nudging has a constant magnitude of 0.4, while the eastward (spring) nudging has a variable magnitude dependent on the depth of the mixed layer. Shallower MLD during the spring migration are related to faster zonal velocities towards the coast, which peak in June when albacore outside of the CCLME experience low and shoaling MLD. A custom fit quadratic equation for the magnitude of the nudging (ranging from 0.1 to 0.6) programs the simulated albacore to travel faster as the MLD shoals further, allowing for a closer match between the seasonal average velocity profile of the simulated and observed tracks. No similar relationship to MLD was found for the westward migration, thereby supporting the use of a constant magnitude.

We worked under the assumption that albacore have some inherent memory and navigational capabilities that lead them between spatially distant foraging grounds and know in which direction to migrate. This is supported by Childers et al., (2011)¹⁴ and Muhling et al., (2022)¹⁸, which find that albacore often return to within 300 km of their coastal origin locations after their seasonal migration to offshore regions. Similar site fidelity and ability to perform trans-oceanic migrations over similar tracks year after year has been observed in many other marine migratory species^3,5,31,63. Because the trans-Pacific albacore migration is in large part oriented east-west, the direction of the nudge towards the CCLME is simplified to be 0^o (east) and the nudge away from the CCLME to be 180^o (west).

Climate sensitivity analysis

Ensemble mean SST and MLD projection data for 2070–2099 under the SSP2-4.5 (“middle of the road”) concentration scenario were originally obtained from the NOAA Climate Change Web Portal, which has since been discontinued (https://psl.noaa.gov/ipcc/, 10.1175/BAMS-D-15-00035.1)⁶⁴. The same data products (CMIP6 ensemble means for sea surface temperature and mixed layer depth) can be found on the NOAA Physical Sciences Laboratory Climate Data Repository (https://psl.noaa.gov/data/CMIP6/). These products include the following models: ACCESS-CM2, BCC-CSM2-MR, CAMS-CSM1-0, CMCC-CM2-SR5, E3SM-1-0, E3SM-1-1, IITM-ESM, INM-CM5-0, KACE-1-0-G, MIROC6, MPI-ESM1-2-HR, MPI-ESM1-2-LR, NorESM2-LM, ACCESS-ESM1-5, CESM2-WACCM, CMCC-ESM2, CanESM5-1, CanESM5, E3SM-1-1-ECA, EC-Earth3-Veg-LR, FIO-ESM-2-0, IPSL-CM6A-LR, KIOST-ESM, MCM-UA-1-0, MRI-ESM2-0, NorESM2-MM, and TaiESM1. Due to the coarser spatial resolution of the CMIP6 multi-model ensemble products compared to the GLORYS dataset used in this study, the projected anomalies were interpolated to 1/12^o resolution. The anomalies were then added to the 2003–2013 GLORYS monthly seasonal climatologies to generate projected SST and MLD fields that maintained monthly resolution.

The full swim model was run on three different future scenarios to explore the sensitivity of simulated albacore to changes in SST and MLD: (1) 2070–2099 SST scenario where only temperature is altered to future conditions, (2) 2070–2099 MLD scenario where only the mixed layer is changed, and (3) 2070–2099 SST & MLD scenario where projections for both variables are used. In all scenarios, the location and productivity of the CCLME as well as the number of simulated albacore are assumed to not be affected by the changes in SST and MLD to allow for comparison with the reference period simulation (2003–2013). Maps showing differences in albacore density between future scenarios and the reference period are produced at 1^o spatial resolution.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

LAD, CE, and CD contributed to all aspects of this study and share first authorship. EDL, BAM, and SSK contributed to the study design. EDL, LAD, CE and CD led data analysis and visualization. EDL, BAM and SSK advised on data analysis and interpretation of results. LAD led the writing of the manuscript with CE and CD and contributions from EDL, BAM, and SSK.

Funding

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Data availability

Data from tags available through the Albacore Archival Tagging Program are not posted publicly at the request of the Albacore Research Foundation, but can be made available upon request to LAD. This study has been conducted using E.U. Copernicus Marine Service Information; GLORYS 12V1 product (https://doi.org/10.48670/moi-00021). This study uses output data from CMIP6 multi-model ensemble products originally accessed from the NOAA’s Climate Projection Web Portal (https://psl.noaa.gov/ipcc; https://doi.org/10.1175/BAMS-D-15-00035.1; Scott et al., 2016). They can now be found on the NOAA Physical Sciences Laboratory Climate Data Repository (https://psl.noaa.gov/data/CMIP6/). The MATLAB code used to analyze the data and produce the figures can be found here: https://github.com/lorenzodavidson-git/Albacore-Tuna-Scientific-Reports.git.

Declarations

Competing interests

The authors declare no competing interests.