Recent declines in salmon body size impact ecosystems and fisheries

The study addresses whether and how body size in Pacific salmon has changed over recent decades, the causes underlying these changes, and their ecological and socioeconomic consequences. Body size is a key biological trait linked to fitness, demography, predator–prey interactions, and human value. Global reports indicate body size declines in many taxa driven by climate change and harvest. For Pacific salmon in Alaska—an important and comparatively intact salmon socio-ecological system—local and Indigenous knowledge suggests declining adult sizes, but comprehensive, multispecies, statewide analyses have been lacking. The authors compile six decades of salmon size and age data across Alaska to quantify the magnitude and consistency of size declines, evaluate potential drivers (climate, ocean competition, harvest), and assess consequences for ecosystems and people.

Prior work links climate change to body size declines across taxa (e.g., Soay sheep, aquatic ectotherms, North American birds) and implicates harvest in reducing body size and age at maturity in marine fishes (e.g., Atlantic cod). Earlier salmon studies documented size and age changes regionally and species-specifically in the North Pacific and Canada, with age truncation implicated particularly in Chinook. Research highlights the demographic and trophic implications of size declines and potential reductions in ecosystem services such as fisheries yield. However, the interplay of multiple stressors (climate, competition at sea, harvest) and their contributions to salmon size trends in Alaska remained unclear, warranting a broad, multispecies, multiregion analysis.

Data: The authors assembled Alaska Department of Fish & Game (ADF&G) age-length (AL) datasets spanning 1957–2018, totaling over 12.5 million individual fish measurements from 1,014 sampling locations across Alaska for Chinook (Oncorhynchus tshawytscha), chum (O. keta), coho (O. kisutch), and sockeye (O. nerka). Samples came primarily from commercial harvest (57%) and escapement monitoring (33%), with additional test, subsistence, and sport harvest projects. Standardized quality assurance excluded implausible ages/lengths, duplicate records, inconsistent measurement types (retained mid-eye to fork), and sparse year-location samples (<10 fish).

Temporal comparisons: For each population (location), mean body length was calculated in two discrete windows: pre-1990 (baseline) and post-2010 (recent). Only populations with ≥3 years of data in each window were included (100 sockeye, 34 Chinook, 32 chum, 13 coho).

Trend modeling: To quantify continuous changes, general additive models (GAMs) were fit to annual mean length per population for each species (data from 1975 onward; populations with ≥5 years: sockeye 276, Chinook 202, chum 183, coho 142). Models included fixed effects for region and population and a nonlinear smoothed effect of year. Alternative GAMs with region-by-year or population-by-year interactions (for populations with ≥20 years of data) assessed spatial variation in trends using AIC.

Age and size-at-age: Similar GAMs modeled mean freshwater age, mean saltwater age, and mean length-at-age (by common age classes) with nonlinear year effects and fixed effects for region and population.

Partitioning contributions: Using the discrete-time chain rule, annual changes in mean length were partitioned into contributions from changes in age structure (proportions at age) versus changes in size-at-age. Proportions were averaged among consecutive years, and contributions were computed per population-year and summarized by region.

Drivers of change: Hierarchical Bayesian models related location-specific mean lengths to environmental covariates and competition indices while estimating a species-specific common nonlinear year effect (residual trend). Covariates included large-scale climate indices (PDO, NPGO, MEI annual and winter), Bering Sea ice cover/retreat, nearshore summer sea surface temperature (ERSST v4 coastal grid cells), regional air temperature (as freshwater proxy), and indices of salmon abundance representing potential competitors (North Pacific and Alaska pink, chum, and sockeye abundances; including hatchery and wild where appropriate). Bayesian R^2 quantified explained variance. Population-specific effects and species-level mean covariate effects were inferred with posterior distributions.

Harvest analysis: For a subset of 33 populations (25 sockeye, 8 Chinook) with sufficient data, linear models tested for association between average harvest rate and long-term change in body size.

Ecosystem and socioeconomic consequences: Species-specific length–weight relationships converted length change to mass change. Per-capita changes were estimated for: (a) fecundity using published Alaska fecundity–length functions; (b) nutrient transport as grams of phosphorus (0.38% of wet weight for spawning adults); (c) commercial value by applying recent region- and species-specific ex-vessel prices to pre-1990 vs post-2010 average weights; and (d) rural food security via edible mass using recovery rates (Chinook 55%, chum 60%, coho 57%, sockeye 53%) to estimate changes in 100 g servings and meals (species-specific meal sizes from Yukon communities). Sensitivity and data source details are provided in Supplementary materials.

Additional considerations: Sexes were not analyzed separately due to unreliable sexing in ocean-phase fish, but prior studies suggest similar trends by sex. Potential fine-scale changes in sampling gear or locations were not modeled due to inconsistent metadata.

• Widespread declines: All four species showed smaller average body sizes post-2010 compared to pre-1990. Mean declines in body length: Chinook -8.0%, coho -3.3%, chum -2.4%, sockeye -2.1%. Magnitude varied by region and population (e.g., Chinook in Westward and Arctic–Yukon–Kuskokwim about -10% vs Southeast -4%).
• Nonlinear temporal patterns: GAMs detected significant nonlinear declines through time in each species (year effect p < 0.0001). R^2 for species-specific GAMs of mean length: Chinook 0.453, sockeye 0.621, coho 0.687, chum 0.784. Declines accelerated after ~2000 across species; sockeye, chum, and coho shared declines beginning mid-1980s, partial recovery early 1990s, then sharp post-2000 decreases. Chinook exhibited a more linear long-term decline with acceleration after 2000.
• Spatial structure: Chinook and coho exhibited high spatial variability best captured by population-specific nonlinear trends; sockeye and chum were better explained by regional-level patterns.
• Age structure vs growth: Shifts in age structure explained on average 88% of interannual variation in mean size. Contributions of changing size-at-age were smaller: coho ~20%, Chinook 7.4%, chum 7.1%, sockeye 5.9%.
• Environmental drivers: Hierarchical Bayesian models explained a moderate fraction of variance (Bayesian R^2: sockeye 0.28, Chinook 0.29, chum 0.35, coho 0.48). Multiple climate and competition metrics had small-to-moderate associations with size. Alaskan pink salmon abundance had a negative association with size across all species (strongest in sockeye). For sockeye, North Pacific pink abundance was particularly negatively associated with body size. For chum, NPGO had a strong negative association, whereas for coho NPGO was positive. Temperature–size patterns did not align with metabolic theory predictions.
• Unexplained common decline: After controlling for covariates, residual species-specific trends still showed overall size declines, suggesting additional unmodeled broad-scale drivers.
• Harvest effects: No significant relationship between average harvest rate and long-term size change across the subset analyzed (R^2 = 0.02, F1,30 = 0.56, p = 0.46).
• Ecosystem and human consequences: Size declines translate into reduced per-capita services. For Chinook, median per-fish reductions were estimated as egg production -15%, commercial value -25%, meals -26%, and nutrient transport (phosphorus) -26%. Other species showed smaller but meaningful reductions. The abstract reports estimated average per-fish reductions for Chinook of eggs -16%, nutrients -28%, value -21%, meals -26%.

The analysis demonstrates pervasive, accelerating declines in salmon body size across Alaska over recent decades, with earlier maturation (younger age-at-return) as the dominant mechanism rather than reduced growth at age. Climate variability and oceanic competition, particularly with abundant pink salmon (wild and hatchery-enhanced), are consistently associated with smaller sizes, though effects differ by species and location and appear temporally non-stationary. The negative sockeye–pink association corroborates prior evidence of interspecific competition. Residual declines after accounting for covariates indicate additional broad-scale factors remain to be identified. Harvest does not appear to be a primary statewide driver of size decline, though it may impact certain populations under specific fisheries and gear selectivity regimes. Declining size has tangible ecological and socioeconomic ramifications, reducing fecundity, nutrient subsidies, commercial value, and food availability, especially acute for Chinook where declines in size coincide with lower abundance. Management implications include recognizing that maintaining body size, not just abundance, is important for sustaining ecosystem functions and salmon–people relationships. Given that pink salmon releases and high ocean abundances are within some management control, tools are needed to evaluate trade-offs of hatchery releases and interspecific competition effects. The capacity of increased abundance in some species (e.g., sockeye, chum) to offset per-capita losses is limited for ecological processes governed by size and under fixed escapement policies.

This study provides a comprehensive, multispecies, statewide assessment showing that Alaska’s Pacific salmon have become smaller over the past six decades, with accelerated declines since 2000. Earlier maturation, rather than slower growth, primarily drives these declines. Climate variability and oceanic competition, notably with pink salmon, are key correlates, while broad harvest effects are not supported at the statewide scale. Declining size likely reduces per-capita fecundity, nutrient transport, commercial value, and food security, with the largest impacts in Chinook salmon. Maintaining body size should be an explicit management objective alongside abundance. Future research should: disentangle maturation schedule changes from marine mortality mechanisms; obtain finer-scale data on predator abundances and size selectivity; assess hatchery selection and competition impacts, including tools to evaluate hatchery release trade-offs; investigate non-stationary climate–size relationships; and extend analyses to longer historical baselines to capture pre-1990 changes.

• Data limitations prevented separating contributions of changing maturation schedules from increased marine mortality; finer-scale marine mortality data are needed.
• Insufficient data to rigorously test hatchery selection effects and predator (e.g., killer whale, salmon shark) abundance and size-selective predation; available predator diet data may not generalize regionally.
• Harvest analysis was limited to 33 populations with adequate data; gear- and fishery-specific size selectivity could vary and was not fully addressed statewide.
• Sex-specific analyses were not conducted due to unreliable sexing in ocean-phase samples; trends are assumed similar between sexes based on prior studies.
• Some fine-scale methodological changes (e.g., gear mesh sizes, sampling locations within watersheds) were not modeled due to inconsistent metadata.
• Commercial samples can include mixed stocks (non-local fish), potentially influencing regional estimates.
• Environmental covariates are proxies and may not capture all relevant physical/biological drivers; residual trends suggest unmodeled factors.
• Nutrient loading calculations excluded juvenile export; social and economic estimates depend on assumed recovery rates and market conditions.
• The pre-1990 vs post-2010 comparison captures a limited historical window; earlier declines likely mean impacts are underestimated.