Seasonal ecosystem linkages contribute to the maintenance of migratory polymorphism in a salmonid population

Abstract

The influence of resource subsidies on animal growth, survival and reproduction is well understood, but their ultimate effects on life history have been less explored. Some wild species have a partially migratory life history, wherein migration is dictated based upon threshold traits regulated in part by the seasonal availability of resources. We conducted a large-scale field manipulation experiment where we provided a terrestrial invertebrate subsidy to red-spotted masu salmon. Individuals in stream reaches that received a subsidy had, on average, a 53% increase in growth rate relative to those in control reaches. This increased growth resulted in a greater proportion of individuals reaching the threshold body size and smolting in the autumn. Consequently, 19–55% of females in subsidized reaches became migratory, whereas 0–14% became migratory in the control reaches. Our findings highlight seasonal ecosystem linkage as a key ecosystem property for maintaining migratory polymorphism in partially migratory animals.

1. Introduction

Processes that promote and maintain variation in life histories are fundamental for the long-term persistence of species and populations [1,2]. Partial migration, where both migratory and resident individuals are maintained within a single population of a given species, has evolved in many animal taxa [3]. Propensity to migrate (or not to migrate) is often heritable in many species and their genetic bases have been elucidated in some species, including salmonid fishes [4–6]. However, the decision to migrate also largely depends on environmental conditions [3], such as resource availability and intraspecific competition, which dictate whether an individual reaches specific threshold traits (e.g. body size and/or lipid stores) [7–9]. In nature, ecosystems are often tightly linked by resource subsidies, meaning resources, such as nutrients, prey and detritus, that flow from one ecosystem to another [10,11]. The resource subsidies can substantially increase resource availability for animals [10,12], thereby relaxing intraspecific competition [13] and affecting life-history traits, such as growth trajectory and age at maturation [14–17]. Despite the ubiquitous nature of these subsidy effects, to our knowledge no studies have assessed how subsidies may influence migratory strategies among animals. This may hinder our ability to fully understand the factors regulating migration polymorphism at landscape scales.

In partially migratory salmonids, maturation of and decision-making regarding migration by individuals are based on threshold traits [18]. Specifically, if an individual reaches maturation threshold in the spring, such as body size and/or lipid reserve, its gonad will develop and the individual will reproduce as a stream resident in autumn. If these thresholds are not reached, resources are instead invested in somatic growth. Following this period, if the individual exceeds the thresholds for migration in autumn, it will undergo smoltification and become a migrant. If not, it will remain as a stream resident until the following spring, when maturation-related decisions are again presented. The seasonality of these life-history decisions suggests that the summer growth rate should directly influence the autumn migration decision in partially migratory salmonids [18–20].

In forested streams, which are typical habitat for juvenile salmonids, terrestrial invertebrate inputs peak during the summer, when riparian forest productivity peaks [21]. By contrast, aquatic invertebrate biomass declines from spring to summer owing to shading by riparian vegetation and the emergence of adult aquatic insects [21]. Previous studies have demonstrated that the terrestrial invertebrate subsidy supports or enhances salmonid growth over the summer months, such that it compensates for the seasonal deficiency in aquatic invertebrate prey [13,21,22]. Here, we hypothesized that seasonally delivered terrestrial invertebrate subsidies would promote the frequency of migration, which could act to maintain migratory polymorphism in partially migratory salmonids inhabiting forested streams (figure 1).

Figure 1.

Hypothesized outcomes of migratory polymorphism in red-spotted masu salmon with terrestrial invertebrate subsidies. Numbers indicate figures that contain supporting data (terr. prey, terrestrial invertebrates; aqua. prey, aquatic invertebrates). The seasonal change in the amount of the subsidy (mealworm + terr. prey) emulates the seasonal dynamics of the terrestrial invertebrate subsidy found in temperate streams surrounded by the natural deciduous forests (see also Material and methods, and figure 2).

To test this, we conducted a large-scale field experiment in which the input rate of terrestrial invertebrates to stream reaches was directly manipulated (figure 1a,b). In red-spotted masu salmon, excluding precocious males, individuals of age 0 that exceed a threshold body size during late August to October become migrants and out-migrate to the ocean in November–December [23,24]. Those that are smaller than the threshold size become stream residents [24]. In this study, we first examined if individual growth rates were higher in stream reaches with a resource subsidy (figure 1c,d). Next, we tested if a larger proportion of individuals reached the threshold body size and became migrants in subsidized streams (figure 1c,d). Overall, we provide, to our knowledge, the first evidence that seasonal ecosystem linkages can be a key ecosystem property to maintain migratory polymorphism in partially migratory animals living in seasonally coupled ecosystems (figure 1e).

2. Material and methods

(a) . Study area and experimental design

Our experiment was conducted from June or July to November of 2020 and 2021. We chose two streams in the upper drainage of the Arida River system (34°5′16.3″ N, 135°31′52.1″ E). Both streams, Murogawa (M) and Furonotani (F), are primary habitat for red-spotted masu salmon in the study area and are located above a large dam (electronic supplementary material, figure S1). Because the dam has isolated these populations from marine habitat over 60 years, the proportion of migratory individuals in the study region is low; only 2% of 866 fish observed in nearby streams (n = 14 streams) were found to be migratory. Therefore, we used captive-bred fish from two hatcheries (Hat-S and Hat-T), in which both migrants and stream residents are maintained.

We established two treatments, a subsidy and a control, each with three replicates, for the reaches in streams M and F, respectively (figure 2). In each stream, the two treatments were randomly assigned to four reaches of length 145 m (± 49 m) that had similar habitat characteristics (width = 3.6 ± 1.3–4.5 ± 1.7 m; depth = 22 ± 8–43 ± 12 cm). All study reaches were separated by erosion-control dams (height: greater than 5 m) at the upstream and downstream ends. On 22 July 2020, young-of-the-year (YOY) captive-bred fish from Hat-T were stocked into four stream reaches in stream M (figure 2a). Fish from Hat-S were released into the same four reaches on 5 July 2021 (figure 2b). In stream F, YOY fish from Hat-S and Hat-T were each stocked separately into two stream reaches (figure 2c). In this manner, the experiment was not conducted with full-factorial design to separate the potential confounding effects of year and hatchery (figure 1). However, YOY fish from both hatcheries were subjected to both subsidy and control treatments across both streams; i.e. in total, there were 12 experimental reach conditions over the 2-year study period (two hatcheries × two treatments × three replicates).

Figure 2.

Experimental design in each stream and year (a–c) and the images and input rate of the terrestrial invertebrate subsidy (d). The total invertebrate inputs (mealworm + natural invertebrates) were within the range of the input rates reported in the natural and secondary forests (see also Material and methods).

The stocking density (0.39–0.74 fish m⁻²) was within the natural range of juvenile fish in the study region (0.01–0.88 fish m⁻²; T. Sato, R. Ueda, S. Noda 2022, unpublished data). We reduced the density of wild conspecifics in the study reaches prior to stocking to reduce density-dependent effects. At the time of stocking, the mean fork length (FL) (± s.d.) of the hatchery fish was 97 ± 6 mm (n = 60) for Hat-S fish and 65 ± 9 mm (n = 167) for Hat-T fish. This was close to or slightly larger than the average FL for wild conspecifics in the study area (66 ± 15–90 ± 4 mm across 10 stream sections; Tanaka et al. [25]). Prior to stocking, the adipose fin of each fish was clipped as an identifying marker for mark-recapture surveys.

We emulated a terrestrial invertebrate subsidy by adding mealworms (larvae of Tenebrio molitor) to the subsidized reaches from July until late October in both years. In each subsidized reach, mealworms were added at a rate of 100 mg m⁻² d⁻¹ using 6–12 automatic fish feeders (W × H × D = 6.8 × 14.9 × 8.7 cm, approximately 100 ml capacity; Eheim Co. Ltd., Deizisau, Germany) per reach (figure 2d). The use of mealworms as an invertebrate subsidy has been detailed previously [15,16]. Natural input rates of terrestrial invertebrates were assessed using a typical protocol involving water-filled pan-traps [21,22] and were relatively low (9.33 ± 8.46 mg m⁻² d⁻¹) in the study reaches. This low mean value is probably owing to previous deforestation and the subsequent establishment of conifer plantations in the study area, which has been shown to reduce terrestrial invertebrate inputs to streams [26]. Consequently, the total invertebrate input rate, i.e. the sum of the mealworm subsidy and natural inputs, was 109.33 ± 8.46 mg m⁻² d⁻¹. This is congruent with peak rates of terrestrial invertebrate inputs to temperate streams bordered by natural and secondary forests (27–444 mg m⁻² d⁻¹; Inoue et al. [26]). We reduced the subsidy rate by 50% every two weeks between late August and October to simulate the seasonal decline in terrestrial subsidies reported from natural streams in the temperate regions [21,22].

(b) . Mark-recapture survey

We conducted mark-recapture surveys in August, October and November of 2020 and 2021. Fish were captured in August when the released fish were expected to be settled into the natural streams, which allowed for growth calculations in subsequent months. Because the migratory decision is made in October in the studied masu salmon [23–25], measuring fish at this time was crucial to calculate the threshold body size. In addition, the studied masu salmon reproduce in October, and thus assessment of the maturation of precocious males should also be made in October. Phenotypes, i.e. whether fish were smolts or residents, were assessed in November, because the red-spotted masu salmon is known to complete smoltification in November [24,25] so that they out-migrate to the ocean in November–December [23]. Precocious males are less likely to become migrant in the following years because previous studies have found almost all migrants were age 1, which meant they have to initiate out-migration in November–December at age 0 [27,28].

During surveys, fish were captured using a two-pass electrofishing method with the units set to 200–250 V pulsed DC (LR-24, Smith-Root, Vancouver, British Columbia, Canada). Captured fish were anaesthetized using approximately 0.03% commercially available clove bud oil (COONA Co. Ltd.; [29]), measured (FL to the nearest 1 mm and weight to the nearest 0.1 g) and individually marked by inserting visible implant elastomers (VIE) (North West Marine Technology Inc. Anacortes, WA, USA) into the head (subcutaneously, over the cranium). Five colours of VIE (red, green, yellow, orange and blue), in combination with four locations on the head, allowed a maximum of 625 unique individual marks. The retention rate of the VIE marks during the experiment was high (greater than 95%), and fish that could not be identified owing to the loss of VIE marks were omitted from the growth calculation. Stomach contents were collected from nine randomly selected fish per treatment and hatchery strain (two treatment × two strain, n = 36 fish in total) using pumping and straining to confirm the influence of the subsidy on consumption. Stomach samples were preserved in 70% ethanol and identified in the laboratory to determine the food sources (mealworms and terrestrial and aquatic invertebrates). The adipose fin-clip was sampled and preserved in 99% ethanol for molecular sex determination. Precocious males were identified based on sperm release following gentle pressing of the abdomen in October. Fish were classified as smolts based on three characteristics [23]: a silvery body colour that obscured the parr marks, a blackish colour at the outer edges of the dorsal fin and a slender body shape relative to resident individuals. Because we conducted the smolt assessment by visual inspection, some uncertainty remained (e.g. some fish expressed ambiguous visual characters). To alleviate the uncertainty, we judged the smoltification independently by three persons. We assigned fish as smolt only if the decisions were consistent with all three persons. Otherwise, we regarded fish as non-smolt (i.e. resident). In this way, we might overlook some future smolts, but the results should be conservative on the main conclusion of the present study (i.e. the subsidy increased the proportion of smolts).

Sex identification in migratory assessments is important because the probability of becoming a migrant can be heavily skewed towards females in salmonids [6], including red-spotted masu salmon [25]. To this end, we euthanized 271 fish for sex determination over the course of the study. We further applied molecular sex determination for 147 individuals to reduce the number of fish euthanized. This molecular approach, using the OtY2 marker, has been used previously for our study species [30].

(c) . Statistical analysis

We could not implement the full-factorial design to separate the potential confounding effects of year and hatchery in the present study. In this situation, we constructed two models for each test; one included the year as a fixed effect, while the other included the hatchery strain. In this way, we could not separate the confounding effects of the two factors, but could test the effect of the subsidy treatment by explicitly considering the effects of year or strain.

We ran generalized linear mixed models (GLMMs) using the R package ‘lme4’ to test whether the invertebrate subsidy increased the total consumption rate of individuals. We included the year or strain and its interaction with the subsidy treatment as other explanatory factors in the GLMM models. The interaction term was excluded if it was not significant. Streams and replicated reaches nested within each stream were included as a random effect. Gamma error distribution and log-link function were used owing to the right-skewed distribution of the response variable; the lognormal distribution was also tested, but did not improve the model fit.

We further tested if the addition of a subsidy, sex, fish density, year or strain, and first-order interactions among these parameters affected the specific growth rate of individuals using GLMMs (gamma error distribution and log-link function) with the same random effect structure described above; the lognormal distribution was also tested, but did not improve the model fit. The interaction term was excluded if it was not significant. The specific growth rate was calculated using the following equation [14]: growth rate = ([m₂/m₁]^1/t − 1) × 100, where m₁ and m₂ are the FLs (mm) of a given individual in August and October, respectively, and t is the time (days) between the FL measurements (i.e. 44–70 days). We found that the increment of body size was negative for 42 out of 343 (12%) fish. We measured the growth rate in a growing season for the immature masu salmon (less reasons for reabsorption of resources), and the negative value was mostly −1 and −3 in minimal. Therefore, we concluded that it resulted from the observation error, instead of biologically meaningful body size reduction, and applied zero growth for fish that had the negative values in the GLMM model.

We considered that a larger proportion of individuals with a larger body size in August might emigrate from the control reaches relative to the subsidized reaches owing to a food deficit, which would complicate inferences made from our study. To test this, we assessed if body size in August, treatment type, year or strain and their first-order interactions affected the probability of recapture after October using a GLMM with a binomial error distribution and a logit-link function, and the same random effect structure described above.

We ran a GLMM with a binomial distribution and a logit-link function to determine if treatment, sex, year or strain and their interaction influenced the probability of smoltification, with the same random effect structure described above. The interaction term was excluded if it was not significant. Where the effect of treatment was significant, we teased apart the mechanism by (i) calculating the threshold body size of smoltification for each hatchery of origin and sex, and (ii) examining if more fish exceeded the threshold size in the subsidy-treatment reaches than in the control reaches. To this end, we ran a GLMM to determine the body size (FL in October) at which the probability of smolting was 50% for each hatchery of origin and used that value as the threshold body size. This analysis was performed separately for each sex because the threshold size differs between sexes in masu salmon [25]. We used the body size in October in those analyses because the migratory decision is made in October in the studied species [23–25]. Precocious resident males were excluded from the analyses because they will not become migrant. All statistical analyses were conducted in R 4.1.2 [31] with an alpha of 0.05. The degrees of freedom for the statistical tests of fixed terms was calculated based on the sample size of each explanatory variable in the GLMMs [32].

3. Results

The total consumption rate of all invertebrate categories (i.e. aquatic and terrestrial invertebrates and mealworms) in the subsidized reaches was, on average, 2.8 times higher than in the control reaches; treatment effect was significant both in the GLMM models with year (GLMM_gamma: treatment effect, t₄ = − 3.31, p < 0.001) or hatchery strain (treatment effect, t₄ = − 3.37, p < 0.001; electronic supplementary material, table S1) as a fixed effect, respectively. The total consumption rate was not significantly associated with year (t₂₉ = 0.40, p > 0.05) and hatchery strain (t₂₉ = − 1.37, p > 0.05) in each of the GLMM models (electronic supplementary material, table S1). Mealworms comprised 87 ± 21% (mean ± 1s.d.) of the total food amount consumed by masu salmon, and the mealworm consumption rate was positively correlated with the total consumption rate in the subsidized reaches (n = 24 fish, r = 0.77, p < 0.0001). Thus, the mealworm subsidy was effective in increasing the total consumption rate.

Individuals in the subsidized reaches grew, on average, 53% faster than those in control reaches (figure 3); treatment effect was significant both in the GLMM models with year (GLMM_gamma: treatment effect, t₅ = − 4.06, p < 0.0001) or hatchery strain (treatment effect, t₅ = − 2.70, p < 0.01; electronic supplementary material, table S2) as a fixed effect, respectively. In the GLMM models, the specific growth rate tended to be higher in control reaches with higher fish density (t₂₂₄ = 3.33, p < 0.001 and t₂₂₄ = 2.86, p < 0.01) and lower in control reaches in 2021 (t₂₂₄ = − 2.05, p < 0.05; electronic supplementary material, table S2). No significant effects relevant to hatchery strain were found (electronic supplementary material, table S2). The probability of recapture in October was not significantly associated with the body size in August, treatment type or their interaction (electronic supplementary material, table S3), although it was higher in 2021 than 2020 (GLMM_binomial: z₆₆₃ = 2.33, p < 0.05) or was lower in Hat-T fish than in Hat-S fish (z₆₆₃ = − 1.96, p < 0.05; electronic supplementary material, table S3).

Figure 3.

Effect of providing a terrestrial subsidy (versus the control) on the specific growth rate of individuals from two hatcheries. Box plots show the mean values calculated across all experimental reaches per treatment (n = 3 per hatchery).

The provision of a terrestrial subsidy significantly increased the probability of smoltification; treatment effect was significant both in the GLMM models with year (GLMM_binomial: treatment effect, z₅ = − 3.65, p < 0.001) or hatchery strain (treatment effect, t₅ = − 3.53, p < 0.01; electronic supplementary material, table S4) as a fixed effect, respectively. Only five out of 254 males became smolts across all reaches and sex was significantly associated with the probability of smoltification both in the GLMM models with year (z₄₀₈ = − 4.47, p < 0.0001) or hatchery strain (treatment effect, t₄₀₈ = −4.35, p < 0.0001; electronic supplementary material, table S4) as a fixed effect, respectively. The probability of smoltification in females increased with increasing body size in October (GLMM_binomial: z₁₉₄ = 6.05, p < 0.0001), and the hatchery of origin had a significant effect (z₁₉₄ = 3.65, p < 0.001; figure 4a; electronic supplementary material, table S5a). Based on the GLMM model, the threshold body size for females was estimated to be 140 mm for Hat-S females and 118 mm for Hat-T females (figure 4a,c). The probability of smoltification in males originating from Hat-S increased with increasing October body size (z₁₄₉ = 2.58, p = 0.01; figure 4b; electronic supplementary material, table S5b), but this was not assessed for Hat-T males as only one Hat-T male became a smolt during the study period (figure 4g). The threshold size for Hat-S males was estimated to be 149 mm, which was approximately 10 mm larger than that of Hat-S females. Overall, 18% and 16% of the females from Hat-S and Hat-T, respectively, reached the threshold body size in the subsidized reaches (figure 4d,f), whereas no females reached this size in the control reaches (figure 4d,f). In total, 19–55% of females became migrants in the subsidized reaches, compared to 0–14% in the control reaches (figure 5). No males from Hat-S exceeded the threshold body size (figure 4e,g), and four of the five male smolts were found in the subsidized reaches.

Figure 4.

Fitted logistic regression lines (with 95% confidence intervals) representing the probability of smoltification based on the FLs of individuals from Hat-S (a,b) and Hat-T (c) in October, and frequency distributions of FLs of individuals from Hat-S (d,e) and Hat-T (f,g) recaptured in October. Vertical lines within each panel represent the threshold body size for each category. A fitted logistic regression line was not available for Hat-T males because only a single male from Hat-T underwent smoltification during the study period. PM, precocious male; IM resident, immature resident.

Figure 5.

Proportions of life-history phenotypes in subsidized and control reaches. Proportions are shown for both sexes and hatchery strains. PM, precocious male; IM resident, immature resident.

4. Discussion

Using a large-scale field experiment, we demonstrated that provision of a terrestrial invertebrate subsidy enhanced the summer growth of red-spotted masu salmon, a partially migratory salmonid species. Enhanced summer growth allowed a greater number of individuals to attain the threshold body size and undergo smoltification during autumn. Thus, a greater proportion of individuals became migrants in the subsidized reaches relative to the control reaches. To the best of our knowledge, this is the first study assessing the pathways by which resource subsidies can maintain migratory polymorphism in seasonally coupled ecosystems.

(a) . What factors make the subsidy effects strong?

We postulate that the effect of terrestrial subsidies on migratory decisions was strengthened by the seasonality of the subsidies and its temporal matching with the migratory decision window in masu salmon, whereas the food availability per se is well known to affect probability of migration in salmonids [33] and other fish species [34]. Under natural conditions, the terrestrial invertebrate subsidy peaks in summer, when aquatic invertebrate biomass is reduced [21]. This seasonal asynchronies in the maxima of biomass between aquatic and terrestrial invertebrates probably make terrestrial prey a major energy source for stream salmonids during the summer months [21,22]. In fact, the terrestrial invertebrates comprised 32–99% of the total food consumed among individuals and increased mean growth rate of fish in the subsidized reaches. The growth rate varied largely among individuals even in the subsidized reaches with varying fish density or potential of among-year environmental fluctuation. Although we cannot provide mechanistic explanations for the growth variation, drifting mealworms (experimentally added prey) might intensify the opportunity for interference competition among individuals [35], which might result in causing the growth variation in uncontrolled environmental situations during the experiment.

A temporal match between the delivery of a subsidy and the migratory decision window should ultimately intensify the effect of the subsidy. Specifically, age 0 male masu salmon of sufficient body size in spring initiate investment in gonad development and become precocious males [24]. Non-precocious males and females invest in somatic growth during the spring-to-summer period and make the migratory decision in autumn; Kuwada et al. [24] experimentally confirmed that fish which attained larger body size in October became smolts from November to December more frequently than smaller conspecifics. We therefore conclude that the terrestrial subsidy provided from July until October in this study functioned as key seasonal prey for masu salmon, leading to increased summer growth and thus allowing a greater proportion of individuals to become migrants, based on greater rates of salmon reaching the threshold body size, in autumn.

We claimed that the importance of temporal match between resource availability and life-history decision windows is robust in animals with a condition-dependent strategy in seasonal environments. However, it also depends on the reaction norm of the population and patterns of natural selection. For instance, if threshold traits are regulated strongly by the environmental factors, such as physiological factor (e.g. temperature) and migration costs [36], the magnitude and seasonal shifts of the subsidy may have relatively weak effect on the migratory decision. Similarly, the effect of the subsidy may become inefficient if the migratory polymorphism is maintained predominantly via the frequency-dependent selection [37], although this is often reported in males in salmonid fishes [38,39].

(b) . Conservation and management implications

Seasonally recurrent resource subsidies are common in natural ecosystems [11,12,40]. However, human-induced environmental changes, including global climate change, are altering the seasonal timing of the subsidies in many ecosystems [41]. If those shifts in the subsidy timing cause the phenological mismatch with the life-history decision windows, they may have more pronounced consequences in the migratory decision and hence the maintenance mechanism of partial migration than we previously thought. Future work that directly tests the importance of relative timing of the resource availability and life-history decisions may offer more nuanced insight into the environmental factors regulating partial migration in changing environments, rather than predictions based on the change in the absolute amounts of resource availability [42,43].

In temperate forested streams, where riparian vegetation consists of deciduous trees, the terrestrial subsidy occurs from spring through to summer [21,22,26], which is seasonally matched with the migratory decision window for the red-spotted masu salmon. However, extensive non-native conifer plantations (Cryptomeria japonica and Chamaecyparis obtusa) have substantially reduced the magnitude of the subsidy in the studied stream, as well as other forested streams [26]. This would cause less individuals to become smolts because of their low summer growth. In addition, the conifer plantation can potentially shift the subsidy timing towards late summer to autumn [44], which might have caused seasonal mismatch with the migratory decision window and hence reduced the frequency of individuals to become migrants. It is important to consider forest management strategies, by which life-history decisions of salmonids, iconic fisheries targets, will not be impaired.

(c) . Rationale for the interpretations of the results

Given that we used captive-bred fish in this study, the results should be carefully interpreted prior to application to wild populations. Migratory strains of red-spotted masu salmon have been produced in some hatcheries; a nearly complete migratory strain (95% of the individuals underwent smoltification and was developed within 12 generations under selective breeding [45]). Therefore, there is some possibility that the proportion of migratory individuals in our study represents an overestimate. However, there was no evidence that the two strains we obtained had been selectively bred for migration, and we note that very few individuals in the control treatments became migratory, suggesting that this effect may be negligible if present.

Captive-bred salmonids often exhibit a reduced growth rate in natural streams relative to wild individuals, probably owing to acclimation to the hatchery environment [46–48]. However, the specific growth rates of fish in subsidized reaches in both streams were very similar to those observed in wild conspecifics used in similar subsidy experiments conducted in nearby streams (electronic supplementary material, supplementary results, figure S2 and table S6). On the other hand, body size at the beginning of the experiment in July was larger for Hat-S fish (mean FL (± 1s.d.) = 97 ± 6 mm) compared to wild conspecifics (66 ± 15–90 ± 4 mm). It is possible that this initial ‘head start’ in hatchery fish may have influenced the results for the Hat-S fish [49,50], although the subsidy produced similar responses in terms of the probability of smoltification in fish from Hat-T that had smaller initial size (65 ± 9 mm).

We suggest that the subsidy enhanced individual summer growth, such that a greater proportion of individuals were able to attain the threshold body size and undergo smoltification. However, an alternative explanation is that individuals with larger initial body sizes and higher possibility to become migrant remained more frequently in the subsidized reaches. In this study, the recapture probabilities in October were not related to body size in August or treatment. While we did not measure the recapture probability from the time at release until August, we previously determined no size dependency for the initial loss of hatchery fish [25]. Those suggest the latter explanation is unlikely.

(d) . Future perspectives

Migratory polymorphism is fundamental to maintaining resilience at the population level in changing environments [51] and ensuring that partially migratory species can adjust to climate change [52,53]. In our study, provision of a subsidy maintained both migratory and resident females at the population level (19–55% of females became migratory), whereas few females (0–14%) became migratory under less-subsidized conditions, representing highly managed artificial forests [26]. In the case of salmonids, migrants often increase the arthropod biomass in streams and riparian habitats [54] by transporting nutrients from marine systems upon their return [55,56]. In turn, this can increase the growth rate of juvenile fishes, which may then affect the outcomes of migratory decisions among that generation [57], creating an eco-evolutionary feedback system [58,59].

A long-standing question in ecology and evolution is how genes and environments synergistically generate and maintain life-history variations [60]. In the era of genomics, the genetic basis of life-history traits, including migratory tendencies, is being elucidated in wild organisms [4–6]. Continued work in this field will enable direct tests of the relative contributions of genes and environmental factors to the expression of life-history strategies in natural systems [61,62]. Our study presents new insight into this field. We emphasize the need to consider environmental factors at the landscape level, where resource subsidies govern seasonal growth and life-history expressions.

Ethics

We conducted the experiment in accordance with the Kobe University Animal Experimentation Regulations (KUAER), which follows the Guidelines for Proper Conduct of Animal Experiments (GPCAE) of the Science Council of Japan. The KUAER and GPCAE do not list fish as laboratory animals (mammalian, avian or reptilian species), and therefore ethical approval was not officially required. However, we did follow the protocols suggested in those guidelines to minimize the pain for the experimental animals (i.e. fish were killed using a high dose of anaesthetic).

Data accessibility

All data is available in Dryad: https://doi.org/10.5061/dryad.sbcc2frbn [63].

Data are also provided in the electronic supplementary material [64].

Authors' contributions

T.T.: conceptualization, data curation, formal analysis, writing—original draft, writing—review and editing; R.U.: conceptualization, investigation, writing—review and editing; T.S.: conceptualization, formal analysis, funding acquisition, investigation, project administration and writing—original draft.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was partially supported by The Asahi Glass Foundation for Environmental Field Research in 2019, the River Fund of The River Foundation, Japan in 2021, and a grant from the Fujiwara Natural History Foundation (2021–2022).