Climate-driven global redistribution of an ocean giant predicts increased threat from shipping

The study addresses how climate-driven redistribution of habitats for highly mobile marine megafauna, specifically whale sharks (Rhincodon typus), will alter overlap with human threats, notably shipping. Prior work shows marine taxa track warming rapidly and shift poleward, to higher latitudes or greater depths, with ocean species moving up to six times faster than terrestrial taxa. Yet, for marine megafauna, monitoring constraints have limited global-scale, dynamic assessments of future habitat locations and their overlap with anthropogenic risks such as ship strikes or fishing. The authors test whether whale sharks conform to common range-shift hypotheses (for example, ocean-basin poleward shifts) and quantify how projected habitat changes affect co-occurrence with global shipping. Given whale sharks’ circumtropical distribution, surface-orientation, and IUCN Endangered status, even modest redistributions could disproportionately elevate collision vulnerability. The work fills a gap by providing quantitative, range-wide projections linking climate-driven habitat change with a key mortality source.

Background literature indicates that marine ectotherms are highly responsive to temperature, tracking isotherms with fewer dispersal barriers than on land. Global projections suggest widespread marine redistributions, with temperate and polar regions acting as sinks and tropical regions as sources. Studies document divergent species responses based on life histories and evidence of habitat losses and displacement among marine predators. Whale sharks are vulnerable to ship collisions due to surface use and overlap with marine traffic. Forecasts also predict substantial increases in shipping capacity by mid-century, potentially compounding collision risk. Despite this, global, movement-informed predictions of future habitat overlap with anthropogenic threats for marine megafauna have been lacking.

Data: The authors compiled a global satellite-tracking dataset of whale sharks (n=348 individuals; >15,000 collective transmission days) from seven ocean regions (North Atlantic, South Atlantic, Northwest Indian Ocean, Southwest Indian Ocean, East Indian Ocean, West Pacific, East Pacific) over 2005–2019. Positions were obtained via Argos Doppler and pop-off satellite archival tags; erroneous or post-2019 locations were removed, yielding 18,745 regularized daily locations. Gaps up to 3 days were interpolated. Species distribution modelling: Presence–background models were built using generalized additive models (GAMs; mgcv bam, fast REML), with a prevalence-standardized presence:background ratio of 1:10. For each presence, 100 randomized points within tag error radii were used to extract environmental data. Background sampling excluded the minimum convex polygon (MCP) of presences within each region and sampled the accessible range between 40°N–40°S. Twenty-eight essential ocean variables (EOVs; dynamic and physical) were extracted spatiotemporally (interpolated, centered, scaled), winsorized as needed. Sex and size were random effects; month used cyclic splines; individuals were random effects in regional models. Eight global hypotheses (surface and subsurface EOV sets) were compared via wAIC; selected variables informed eight regional hypotheses. Surface-only models were ultimately used for global coverage of shallow, nearshore habitats. Algorithm control included BART models (200 trees) for comparison; agreement maps between GAM and BART were produced. Data thinning and validation: To reduce autocorrelation, locations ≥2 days apart were retained. Internal 10-fold cross-validation assessed accuracy, precision, sensitivity, specificity, AUC, kappa, and TSS. External validation used OBIS encounters (n≈9,379) and verified Sharkbook.ai observations (n≈13,437), evaluated with continuous Boyce index. Biological realism was qualitatively checked against known seasonal distributions and expert knowledge across validation regions. Climate projections: CMIP6 EOV projections for mid-century (2046–2055; “2050”) and end-century (2086–2095; “2100”) were used under SSP126 (sustainable development), SSP370, and SSP585 (high emissions). GCMs included ACCESS-ESM1-5, CanESM5, CESM2-WACCM, CMCC-ESM2, GFDL-ESM4, IPSL-CM6A-LR, MPI-ESM1-2-HR, and NorESM2-MM. A delta-change framework created debiased ensemble EOV layers for each decade–scenario combination. Regional GAMs were projected basin-wide within known latitudinal limits. Core habitat was defined as the 90th percentile of habitat suitability within each map; change in core area and latitudinal shifts were calculated (also explored at 50th, 75th, 95th percentiles). Shift rates were computed for northerly and southerly edges. Shipping co-occurrence: A ship co-occurrence index (SCI) adapted from a validated collision risk index used 2019 Global Fishing Watch monthly average shipping density (vessels >300 GT; 0.2° grid) overlaid with whale shark habitat suitability (0–1). SCI was averaged within EEZ marine regions (n=367) for baseline (2005–2019) and future decades/scenarios; percentage change from baseline was computed. Analyses also considered effects when holding ship numbers at 2019 levels versus projected fleet growth (up to ~1,200% by 2050 referenced in literature). SCI represents relative co-occurrence, not absolute risk, as dynamic movement and behavioral depth use were not modeled. Geopolitical and ecosystem units: Changes in mean suitability and core habitat coverage were summarized within Exclusive Economic Zones (EEZs; n≈200 for some summaries) and Large Marine Ecosystems (LMEs), examining intra-annual (seasonal) and interannual trends.

- Global redistribution: Models project increased habitat suitability at current range edges and a redistribution from present-day centers toward fringing habitats. Medium-importance LMEs (current mean suitability 0.05–0.5) generally become more suitable, while high-importance LMEs (>0.5) become less suitable by 2050–2100, with magnitude dependent on scenario (Kruskal–Wallis χ²=42.30, P≈5.1×10⁻¹⁰ for n=28 LMEs). - Core habitat change: By 2050, core habitat (90th percentile) decreases in the East Pacific, East Indian Ocean, and South Atlantic, and increases in the West Pacific, Southwest Indian Ocean, Northwest Indian Ocean, and North Atlantic across scenarios. By 2100, differences amplify, with >5 million km² changes in core area in some regions (e.g., North Atlantic gains; East Pacific losses). - Latitudinal shifts: Northerly core habitat cold edges shift at ~12 km yr⁻¹ overall (mid-century ~15 km yr⁻¹; end-century ~9 km yr⁻¹), about 2.5× faster than southerly cold edges (~5 km yr⁻¹ overall; mid ~6; end ~3), consistent with greater Northern Hemisphere warming. West Pacific shows pronounced northward core displacement (~1,300 km by 2050 even under SSP126). Poleward extensions occur broadly, but region-specific patterns include north Atlantic shifts away from Gulf of Mexico and increased equatorial suitability. - EEZ-scale impacts: Under SSP585 by 2100, 57.5% of EEZs experience >50% loss in suitable habitat and 76.5% see >50% reductions in core habitat coverage (n=200). Losses are most prevalent in Asia (88.0% of Asian-sovereign EEZs) and least in Europe (42.1%). Under SSP126, 65.5% of EEZs gain >50% core habitat coverage; Europe shows the greatest gains (73.7%). - Seasonal shifts: LMEs show changing seasonal suitability patterns (e.g., Guinea Current LME exhibits expanded/strengthened Nov–Mar season by 2100; Southeast US Continental Shelf LME shows contracted/weakening season under SSP585). - Mechanistic context: East Pacific future oligotrophication (warming, shallower deep scattering layers, reduced chlorophyll a) may render large equatorial upwelling areas unsuitable. North Atlantic projections reflect warming/salinization and expansion of low-productivity subtropical gyre conditions. - Shipping co-occurrence: SCI increases across all future decade–scenario combinations, even with ship numbers fixed at 2019 levels. Excluding EEZs with mean SCI <0.1 in both periods (n=295), mean EEZ change indicates SCI is ~15,000× greater by 2100 under SSP585 relative to baseline; under SSP126, increases average around 20×. Averaged across EEZs, SCI rises by ~41.2% by 2100 (SSP585) and ~19.2% (SSP126). Regional examples: US North Pacific EEZ SCI increases by factor ~95 due to new suitable habitats overlapping busy routes; Japanese eastern China Sea +272%; Sierra Leone +689%; Somali EEZ +236% driven by offshore suitability where shipping remains low. Decreases occur where suitability contracts away from busy routes (e.g., −76% in Mexican Gulf of Mexico; near −100% in Clipperton region due to general habitat loss).

The results support and refine climate-driven redistribution hypotheses for marine megafauna by showing strong poleward and range-edge increases in habitat suitability for whale sharks, yet with region-specific deviations driven by local oceanography (for example, gyre expansion, oligotrophication). Critically, the projected redistribution substantially elevates co-occurrence with global shipping, implying higher collision exposure for a surface-oriented, Endangered species. Even with static shipping effort, SCI increases markedly, and under high emissions, co-occurrence escalates dramatically. These findings directly address the study’s question by quantifying not only where habitats will move but also how overlap with a major mortality source will change. Conservation relevance is high: expected core habitat losses in some national waters, shifts in seasonal windows, and growing co-occurrence with ships suggest potential population-level impacts (e.g., altered access to aggregation/foraging sites, changes in breeding/birthing locales). Integrating climate-threat projections into management is essential, including spatial planning that anticipates future habitats, evaluating MPA resilience to climate change, and developing measures to mitigate ship strikes. Incorporating the vertical dimension (diving behavior) and dynamic movement into future risk models is highlighted as an important next step.

This study provides the first global, movement-informed projections of whale shark habitat redistribution under multiple climate scenarios and links these changes to co-occurrence with shipping. It shows substantial poleward shifts (>1,000 km), widespread redistribution from core regions to range edges, major EEZ-scale core habitat losses under high emissions, and steep increases in ship co-occurrence—especially under SSP585. The methods (satellite tracking, GAM/BART SDMs, CMIP6 delta-ensemble, and SCI overlay) are transferable to other marine megafauna to inform climate-smart conservation planning. Future work should: (1) incorporate vertical habitat use and behavioral responses to refine collision risk; (2) include projections of changing shipping routes and effort; (3) assess compound stressors (e.g., heatwaves, deoxygenation) and cumulative impacts; and (4) optimize adaptive, multinational protection networks encompassing future aggregations, hotspots, and refugia.

- Regional specificity and overextrapolation: Predictions beyond tracking-informed regional boundaries may overextrapolate and should be interpreted cautiously; localized conditions can drive deviations from global trends. - Model structure and variables: Final projections used surface EOVs to ensure shallow/coastal coverage; some subsurface dynamics and prey proxies (e.g., epipelagic micronekton) were excluded due to uncertainty and lack of future projections. - Presence–background design: Results depend on background sampling assumptions (MCP-exclusion accessible area); method suitability may vary by species. Prevalence effects were managed but can influence probability outputs. - Validation limits: Despite internal and external validations, model performance varies among regions and algorithms; uncertainties remain in absolute habitat suitability magnitudes. - Shipping co-occurrence (SCI): SCI estimates static co-occurrence using 2019 shipping density and do not incorporate dynamic movements, behavioral depth use, or future rerouting/effort changes. Thus, SCI reflects relative overlap, not absolute collision risk. - Climate projection uncertainty: Use of multi-model CMIP6 ensembles mitigates single-model bias, but scenario and model uncertainties persist, particularly for regional ocean biogeochemistry (e.g., chlorophyll, salinity).