Contrasting patterns of population structure and breeding dispersal exhibited between sister taxa in a small river system: new insights from population genetics and sibling analysis on Atlantic salmon (Salmo salar) and brown trout (Salmo trutta)

Abstract

Background

The movement of individuals from one breeding site to another (i.e. breeding dispersal) can have important consequences for fitness and population genetics. Whilst closely related species are expected to be phenotypically similar in many respects, they can nevertheless exhibit important differences in behaviour and ecology that can have repercussions for population structure.

Methods

Here we use genetic information (microsatellite data) collected on two sister species – Atlantic salmon (Salmo salar) and brown trout (Salmo trutta) – that overlap in their spawning and freshwater rearing habitats within a small (ca. 100 km²) catchment in western Ireland. We first explore differences in the spatial extent of within-season breeding dispersal. To do so, we combine sibship reconstruction analysis with distances between siblings’ sampling locations and introduce the ratio of half-sibling to full-sibling pair distances as a novel proxy measure for breeding dispersal (I_BD). We then use a range of population genetic analyses to assess the genetic structure of both species.

Results

The I_BD value was 1.8 times higher in Atlantic salmon than in brown trout, and distances between breeding sites were much greater in salmon (typically within 4 kilometres) than in trout (typically within 600 metres). No genetic structure was detected in Atlantic salmon. In contrast, fine-scale genetic structuring was observed in the brown trout data with at least two genetic clusters being detected, confirming similar patterns reported in previous studies.

Conclusions

Our results support the hypothesis that Atlantic salmon engage in more spatially extensive breeding dispersal behaviour than brown trout, which in turn may explain why trout but not salmon exhibit population genetic structure in this system. Collectively, our findings showcase how the same genetic data can be used to gain complementary insights into the movement behaviours of individuals and the genetic structuring of populations, with implications for conservation management. They furthermore emphasise a hitherto overlooked role for intra-seasonal breeding dispersal in shaping spatial genetic connectivity in polygamous species.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s40462-026-00678-2.

Background

The movement of individuals from their birthplace to their breeding site (natal dispersal) or between sites across multiple breeding events (breeding dispersal) varies across living organisms in scale, strategy, and behavioural characteristics, and is shaped by a wide range of genetic, ecological, and environmental factors [1–8]. Dispersal capacity within populations, and its variation through time, drives patterns of gene flow and can generate genetic differentiation, with local adaptation potentially contributing to but not required for population structuring [9, 10]. Even closely related species can display contrasting dispersal capacities and gene flow patterns. For instance, Engler et al. [11] showed that three sister species of the butterfly genus Thymelicus exhibit dissimilar dispersal abilities, which were associated with differences in both the pattern and scale of genetic differentiation. In particular, the two species T. acteon and T. lineola illustrate this contrast: the former is sedentary and restricted to xerothermic calcareous grasslands, showing effects of land-use patterns on genetic differentiation, whereas the latter occupies a much broader ecological niche and exhibits strong dispersal behaviour, resulting in a more homogeneous genetic structure at the same spatial scale. It is not surprising that such differences are observed, as close phylogenetic relatedness does not necessarily imply close similarity in behaviour or ecological traits [12]. Such examples are also reported in fish. For example, Baek et al. [13] showed that different life-history strategies in the sister sculpin species Cottus hangiongensis and C. koreanus affected population connectivity and, consequently, genetic structure. The former is amphidromous (migrating between freshwater and the sea for purposes other than breeding) and exhibits a homogeneous genetic population, whereas the latter is strictly fluvial and displays genetically discrete populations. These contrasting patterns of genetic structure are most likely explained by the greater larval dispersal capacity of Cottus hangiongensis. Regarding breeding dispersal in particular, no study to our knowledge has explicitly demonstrated parallel contrasting patterns in both breeding dispersal and genetic structure between closely related species. However, evidence suggests that breeding dispersal patterns can differ substantially among closely related taxa in relation to factors such as the mating system. For example [14], showed that breeding site fidelity (used as an inverse proxy for between-season breeding dispersal) varies markedly among shorebird species. In particular, three closely related species of similar body size that breed at comparable latitudes—the great knot (Calidris tenuirostris), the ruff (C. pugnax), and the pectoral sandpiper (C. melanotos)—exhibit striking differences in site fidelity, with return rates of approximately 92%, 33%, and 1.5%, respectively. The authors attribute these differences primarily to variation in mating systems, with the monogamous great knot contrasting with the polygynous ruff and pectoral sandpiper, which exhibit female-only parental care.

Atlantic salmon (Salmo salar) and brown trout (Salmo trutta) are genetically closely related species, and hybridization events between them are frequently observed [15]. They share several ecological and behavioural traits, including similar diets and the ability to adopt alternative life-history tactics, such as anadromy (i.e., migration from freshwater to the sea to grow and mature and return from the sea to freshwater to breed) or remaining in freshwater to complete their entire life cycle [16, 17]. However, they also display important differences. For instance, anadromy is the most common life-history strategy in Atlantic salmon, whereas fully freshwater life histories (e.g. river residency, migrations within freshwater) are more common in brown trout. In addition, anadromous Atlantic salmon typically migrate directly to the open ocean, with only rare populations remaining in coastal or brackish waters, while anadromous brown trout usually remain in coastal areas and generally spend much shorter periods at sea [17, 18]. These contrasting migratory strategies lead to further differences, including the scale and pattern of genetic structure. Atlantic salmon frequently exhibit detectable genetic structure within large river systems (e. g. >1000 km²), as reported in several studies [19–22]. However, there is a relative lack of studies examining genetic structure for Atlantic salmon in smaller river systems. Genetic structuring in small catchments is an important management consideration, as approximately 60% of the ~144 Irish river systems recognised as supporting salmon populations and assessed annually for catch quota provision [23, 24], have a catchment area of approximately 100 km² or less [25]. In contrast, brown trout exhibit substantial genetic structure not only across large catchments [26–28]); but also at much finer spatial scales [29–32].

In both Atlantic salmon and brown trout, breeding dispersal may occur either within a single breeding season (fractional spawning) or across seasons (iteroparity) and may constitute an important component of overall dispersal (which additionally includes natal dispersal). However, little is known about the extent to which breeding dispersal drives gene flow in these species, and the spatial scales at which this occurs. In Atlantic salmon, Taggart et al. [33] reported in the Girnock Burn (a tributary of the Aberdeenshire Dee in Scotland) that some individuals spawned over distances occasionally exceeding 5 km, while no direct estimates of the spatial extent of multiple within-season spawning events by the same individuals available for brown trout, to the best of our knowledge. Moreover, no study to date has compared this phenomenon for both species living in sympatry within the same freshwater catchment. The fact that genetic structure has been shown to occur at much finer spatial scales in trout than in salmon suggests that differences in dispersal behaviour—potentially including breeding dispersal, in combination with natal dispersal—may contribute to higher levels of gene flow in salmon across larger spatial scales.

Several studies have used sibship reconstruction to investigate dispersal patterns in natural populations, such as intergenerational dispersal in the mosquito Aedes aegypti [34], queen dispersal in bumblebee species Bombus pascuorum and B. lapidaries [35], and individual dispersal in the coral reef fish Elacatinus lori [36]. In salmonids, Hudy et al. [37] demonstrated limited post-emergence juvenile dispersal of brook trout (Salvelinus fontinalis) from their natal redds, revealing fine-scale within-stream spatial population structure. However, to our knowledge, sibship reconstruction has not yet been used to explicitly quantify breeding dispersal.

Here, we used sibship reconstruction analysis to quantify the extent of breeding dispersal in two co-occurring sister taxa—anadromous Atlantic salmon and a predominantly non-anadromous brown trout population—by analysing the distribution of distances separating full- and half-sibling dyads, sampled at the fry/parr life stage, across the river network of the Burrishoole catchment in western Ireland (ca. 100 km²). In parallel, we investigated the genetic structure of both species using admixture inference, Principal Component Analysis (PCA), and estimates of pairwise F_ST among sampling sites across the catchment. This allowed for an evaluation of whether differences in within-season breeding dispersal between the species could be a factor contributing to their contrasting genetic structures by influencing the spatial scale of intra-catchment gene flow (see Box 1).

Box 1.

Estimating the impact of breeding dispersal as a driver of gene flow

Within-season, intra-catchment breeding dispersal is a potentially major source of gene flow in Atlantic salmon and brown trout populations. Because both males and females are polygamous, they can potentially reproduce with multiple partners and at multiple locations. This behaviour promotes local genetic mixing, as well as intra-catchment (e.g. inter-tributary) gene flow when individuals move between sites to breed. The extent of breeding dispersal among multi-spawning individuals can be estimated from the distances between juvenile sampling locations. For example, consider a male that first breeds with a female at a first site, producing a full-sibling family (Fig. 1). If the same male then moves within the river system and breeds with a second female at another site, he produces a second full-sibling family. Offspring from these two families are therefore half-siblings, and the geographic distance between their respective natal sites represents the exact breeding dispersal distance of the male. After hatching, juvenile fish diffuse from their natal sites, generally moving downstream [16], and each juvenile travels a different distance from its point of origin (Fig. 1). Using a genetic dataset of juveniles sampled across multiple sites of known coordinates within a river system, full-sibling and half-sibling relationships can be inferred for all individuals, allowing distances to be calculated for every full-sib and half-sib dyad (Fig. 1). For a full-sibling dyad, the distance between the two juveniles (at the time of sampling) arises solely from their individual post-hatching movements away from the same natal site, assuming that multiple spawning by the same parental pair at different sites is negligible. This distance may be zero if both juveniles settle in the same sampling site, or positive if they distribute to different sites. For a half-sibling dyad, the distance between juveniles can originate from two processes: 1. juvenile movement alone, when the shared parent spawned multiple times at the same site, or 2. juvenile movement plus breeding dispersal, when the shared parent spawned at different sites. In this second case, the additional component corresponds to the breeding dispersal distance of that parent. Therefore, comparing the distribution of distances for half-sibling dyads with the distribution for full-sibling dyads provides a way to isolate and quantify the contribution of breeding dispersal (IBD). Fig. 1. Breeding dispersal and juvenile movement in Atlantic salmon and brown trout, and indirect estimation of breeding dispersal distances using sibship and geographic data

The specific aims of the study were to (i) quantify within-season breeding dispersal in Atlantic salmon and brown trout in the Burrishoole river system and assess how it relates to potential gene flow across the river system; and (ii) characterise and compare the genetic structure of both species in the light of findings from the breeding dispersal study.

Methods

Study system

The Burrishoole River system (Fig. 2) is located in western Ireland (53° 55’ N, 09° 34’ W), covering a freshwater catchment area of ca. 100 km² and has a total productive wetted area of 4.74 km², including 4.49 km² of lacustrine and 0.25 km² of fluvial habitat. It is drained by 45 km of small, shallow rivers and streams and includes three main lakes: Bunaveela (0.46 km²), Feeagh (4.1 km²), and Furnace (1.41 km²), the latter being a brackish, partially tidal lagoon that connects to Clew Bay and the Atlantic Ocean via a short estuary. The region is predominantly covered by blanket bog, with forestry and hillside subsistence farming being the primary land uses. The climate is moderate, characterized by mild winters and relatively cool summers [38]. Potential spawning habitats for Atlantic salmon were identified and mapped onto GIS (Fig. 2) as part of a GPS assisted survey conducted by the Salmon Research Agency of Ireland [39]. Spawning habitat availability is not shown in Fig. 2 for brown trout because it is believed to be much more evenly distributed across the catchment, as trout are less constrained by strict habitat requirements and can spawn across a wide range of water depths, from shallow areas to deep-water habitats [40].

Fig. 2.

Map of the Burrishoole catchment showing sampling sites for Atlantic salmon and brown trout. Dark blue circles indicate Atlantic salmon sampling sites only, purple squares indicate sites where both Atlantic salmon and brown trout were sampled, and red triangles indicate brown trout sampling sites only. Light blue areas represent available spawning habitat for Atlantic salmon. No information was available on the distribution of spawning areas for brown trout, but suitable spawning habitat for that species is believed to be widely distributed across most of the whole catchment. Rough= Rough river, Fid_H= Fiddaunnahoilean river, Goul= Goulaun river, Maum= Maumeratty river, Fid_V= Fiddaunveela river, Lodg= Lodge river, Glen= Glenamong river, Alta = Altahoney

Sampling approach and preparation of genetic material

From 07/08/2013 to 25/09/2013, 530 salmon 1+ parr were sampled in the Burrishoole catchment across 22 sites distributed among eight tributaries (Fig. 2). Although these watercourses are referred to by their official names as “rivers”, they are in practice small tributaries within a single dendritic river network. We retain this nomenclature for consistency with existing maps and datasets; accordingly, the term “river” is used here to refer to individual tributaries of the same main river system. The sampling sites were chosen to approximate the distribution of spawning habitats across the river network [39]. The sampling process consisted of capturing fish by backpack electrofishing. A small caudal fin clip was taken from each sampled individual in order to provide tissue for DNA extraction and genetic analysis. The fish were genotyped at 13 microsatellite biallelic loci: SSsp2215, Ssa171, SSsp2216, Ssa85, Ssa197, Ssa412, EST107, Ssa407, EST28, EST41, EST68, EST123, and EST19 [41]. An identity analysis conducted with CERVUS [42] detected no technical duplicates (e.g. same fish sampled twice) in the dataset. Trout were sampled using the same methods from 2017 to 2020 in Gaulaun, Fiddaunveela, and the Rough River (9 sites; Fig. 2), and genotyped at 16 biallelic microsatellite loci: Ssa416, One103, Cocl.Lav.4, One9uASC, CA048828, One108, Ssa85, One102.a, One102.b, CA053293, ppStr3, CA060177, Ssa197, SsaD71, SaSaTAP2A, and Ssa410UOS [29, 43]. To obtain a trout dataset comparable to the salmon dataset in terms of sampling period and sample size, we retained only individuals collected between 15/08/2018 and 04/09/2018. Using the Mclust package in R [44], individuals were assigned to three size-based clusters, assumed to correspond to age classes 0+, 1+, and 2+ (or older). Only 0+ (34–86 mm) and 1+ (87–137 mm) individuals were retained for subsequent analyses (nₜₒₜ = 360; n₀₊ = 285; n₁₊ = 75).

Assessing the shortest waterway distance between sampling sites

Shortest waterway distances between sampling sites were calculated in QGIS (version 3.30.1–‘s-Hertogenbosch) using a river network layer and point coordinates (coordinate reference system: EPSG:29902). The OD-Matrix from Points as CSV tool from the QNEAT3 plugin (version 1.0.6) was employed, with the sampling site coordinates used as the input point layer and the WATER_RivNetRoutesdataset from the EPA Ireland catalogue used as the network layer. Because the sampling points represented actual locations along the river network, the “network cost” values from the output were taken as estimates of shortest waterway distance for subsequent analysis.

Detection of siblings and dispersal analysis

Full-sibling families and half-sibling dyads were identified using sibship likelihood analysis in COLONY [45]. Three independent runs were performed with the following settings: allele frequencies not updated, inbreeding allowed, both male and female parents assumed to be polygamous, runs set to long, full-likelihood method, and precision set to high. Following the analysis, only full-sibling families and half-sibling dyads with a probability of inclusion (Prob(Inc)) greater than 0.8 were retained. Prob(Inc) represents the likelihood that all individuals assigned to a family are true siblings. For trout, 43 sibship links were detected across different year classes (Supplementary Table 1) and were removed from the analysis to ensure that comparisons reflected within-season breeding dispersal only, consistent with the approach used for salmon. Pairwise waterway distances were subsequently assigned to all full-sibling and half-sibling dyads, taking a value of zero for dyads originating from the same site and the corresponding inter-site waterway distance for dyads originating from different sites. The resulting distance distributions were then generated for each species and, for trout, separately by age class.

On the assumption that greater spatial dispersion of half-siblings relative to full siblings indicates that the parents bred with different partners in different breeding sites (Box 1), an index was developed to characterise the extent of this within-season, within-catchment breeding dispersal for each species separately. To derive this index, at each iteration (10,000 iterations), dyads were first subsampled to reduce non-independence, with one dyad randomly selected per full-sib family and half-sib dyads sampled such that no individual occurred in more than one dyad. One full-sib distance and one half-sib distance were then randomly drawn from these subsets, and the ratio of half-sib to full-sib distances was calculated (with one added to all distance values to avoid division by zero). This procedure yielded a distribution of ratio values, from which the median was taken as the breeding dispersal index . with  = intra half-sibling dyad distance,  = intra full-sibling dyad distance, b  = bootstrap iteration index and B the total number of bootstrap iterations. An I_BD value of 1 indicates no detectable contribution of parental breeding dispersal to offspring spatial distribution, whereas values greater than 1 quantify the relative contribution of breeding dispersal compared to juvenile movement, with increasing values indicating a stronger effect. For instance, a value of 2 means that juveniles are twice as dispersed under the combined effect of breeding dispersal and juvenile movement as they would be under juvenile movement alone (see Box 1). To estimate uncertainty around the final estimate, the resulting bootstrap ratio distributions were subsequently resampled with replacement over 10,000 iterations, and the median recalculated across iterations, generating bootstrap confidence intervals around the median value for each species.

Moreover, assuming that juvenile movement does not differ between individuals in half-sibling and full-sibling dyads (Box 1), the distribution of breeding dispersal (D_BD) was estimated by subtracting the component attributable to juvenile movement from the spatial dispersion of half-siblings. This was achieved using the same procedure, whereby at each iteration one full-sib and one half-sib distance were randomly drawn from subsampled dyad sets, and the difference between half-sib and full-sib distances was calculated, yielding a distribution defined as . For trout, dyad distances from both age classes were pooled to estimate I_BD and the distribution of breeding dispersal. This is not expected to bias comparisons with salmon, as the method removes post-emergence juvenile movement and differences among age classes should not bias the breeding dispersal estimate itself, which is based on relative difference between distances rather than on raw distances.

Assessing genetic structure

Both trout and salmon datasets were filtered to retain a single representative per full-sib family identified in the sibship likelihood analysis for downstream population genetic analyses. This filtering step was applied to reduce the potential influence of family structure on clustering and differentiation analyses, while recognising that removal of siblings can also reduce statistical power and affect parameter estimation [46]. Genetic differentiation between sites was estimated using the F_ST index. Pairwise F_ST values were only calculated between sites with a sample size of at least 25 individuals, in accordance with the recommended threshold for reliable allele frequency estimation [47]. For salmon, a total of 233 individuals were retained for the analysis, including 36 from Glen_03, 40 from Goul_04, 34 from Goul_06, 31 from Maum_06, and 92 from Rough_12. For trout, a total of 265 individuals were retained, comprising 38 from Fid_V_01, 42 from Fid_V_02, 129 from Goul_01, 28 from Rough_04, and 28 from Rough_05. Confidence intervals at 95% were obtained by bootstrapping with one hundred iterations, using default values of the boot.ppfst function from the hierfstat package [48]. To test for correlation between genetic and waterway distances, Mantel tests were conducted using the ade4 package [49] in R [44] (version 4.4.0). Standardized F_ST (F_ST/(1 − F_ST)) was used as a measure of genetic distance, as it rescales genetic differentiation relative to its theoretical maximum, thereby making it independent of marker variability and differences in heterozygosity [50]. Admixture analysis was conducted using PopCluster (version 1.1.0.0 [51]). Parameters were set at medium scaling, with the minimum number of clusters set to one and the maximum set to nine, 10 replicate runs per cluster value, population data available, and other parameters set to default. The best number of clusters was determined using two methods implemented in PopCluster: the DLK2 and FSTIS estimators. Principal Component Analyses (PCA) were conducted in R using the adegenet [52] and ade4 packages. Only sampling locations with at least 25 individuals were retained, consistent with the filtering applied for pairwise F_ST analyses. Genotype data were first transformed using the scaleGen function in adegenet, with missing values imputed by the mean allele frequency, and the data centred but not scaled. PCA was then performed using the dudi.pca function in ade4 (center = FALSE, scale = FALSE), retaining the first 20 principal components. The proportion of variance explained by each principal component was calculated from the eigenvalues.

All genetic structure analyses for trout were also repeated using a subset of the 13 least polymorphic markers, to ensure that comparisons with salmon were not biased by differences in the number of markers used (i.e. 16 vs 13 markers).

Results

Breeding dispersal

After COLONY analysis (n = 530 salmon, 285 age 0+ trout and 75 age 1+ trout) and subsequent filtering, 245 full-sib dyads and 254 half-sib dyads were identified in salmon. In trout, 156 full-sib dyads (150 age 0+ and 6 age 1+) and 121 half-sib dyads (99 age 0+ and 22 age 1+) were detected (Table 1; Supplementary Table 1). Full-sib analysis identified 54 families in salmon and 36 in trout, including 30 in age 0+ and 6 in age 1+, with mean family sizes of 2.80 individuals in salmon, 3.14 in trout 0+, and 2.00 in trout 1+ (Supplementary Table 2). In salmon, family sizes ranged from two to 16 individuals, whereas in trout they ranged from two to 11 individuals. In salmon, 18 families were detected at a single site, 27 across two sites, and nine across three sites, indicating that most full-sib families were spatially distributed across multiple locations. In trout, 21 age-0+ families were sampled at a single site, five across two sites, two across three sites, and one across five sites, whereas among age-1+ fish, two families were confined to a single site and four spanned two sites (Supplementary Table 1). For half-siblings, considering only individuals involved in at least one half-sib relationship, the number of half-sib events per individual in salmon ranged from one (150 individuals) to eight (two individuals), with a mean of 2.07. In trout, the number of half-sib events per individual ranged from one (115 individuals) to five (three individuals) in age 0+ juveniles (mean = 1.58), and from one (19 individuals) to six (one individual) in age 1+ juveniles (mean = 1.81) (Supplementary Table 3).

Table 1.

Summary statistics of pairwise waterway distances (m) between full-sib and half-sib dyads and bootstrapped estimates of breeding dispersal in Atlantic salmon and brown trout

Group	n	Min.	25th percentile	Median	Mean	75th percentile	Max.
Full-sib_salmon	245	0	0	0	1,238	1,911	9,540
Half-sib_salmon	254	0	0	2,337	3,131	5,619	12,100
Breeding dispersal estimation_salmon	10,000	0	0	680	2,146	4,110	12,100
Full-sib_trout (0+)	150	0	0	0	38.2	0	630
Full-sib_trout (1+)	6	0	0	0	90.6	0	544
Half-sib_trout (0+)	99	0	0	0	641	301	10,043
Half-sib_trout (1+)	22	0	0	538	473	669	1,255
Breeding dispersal estimation_trout	10,000	0	0	0	846	538	10,043

Patterns in sibling-distance distributions reveal a much broader spatial spread in salmon than in trout (Fig. 3A–C; Table 1). Both full-sib and half-sib dyads occur over substantially greater distances in salmon, whereas in trout most dyads are confined to the same site or nearby locations (Table 1). Regarding differences in the impact of breeding dispersal between species, the breeding dispersal index (I_BD) is significantly higher in salmon (1.77, 95% CI: 1.59–1.80) than in trout (1.00, 95% CI: 1.00–1.00), indicating a greater relative contribution of breeding dispersal in salmon. The trout confidence interval remains fixed at 1, indicating that half-sib and full-sib dyads very frequently show similar spatial distances in trout, with no detectable over-dispersion of half-sib dyads relative to full-sib dyads, contrary to salmon. Consistent with this pattern, the estimated breeding dispersal distribution (D_BD) is also markedly broader in salmon than in trout (Fig. 3D; Table 1). In salmon, 75% of multi-spawners are estimated to disperse within 4,110 m, compared with only 538 m in trout, highlighting a much greater spatial extent of breeding dispersal in salmon.

Fig. 3.

A: distribution of intra-dyad distances in Atlantic salmon. B: distribution of intra-dyad distances in 0+ brown trout. C: distribution of intra-dyad distances in 1+ brown trout. D: estimated distribution of breeding dispersal (D_BD) in Atlantic salmon and brown trout (half-sib distances minus full-sib distances)

Genetic structure

PCA reveals no clear clustering in Atlantic salmon, with individuals from all sampling locations largely overlapping (Fig. 4A). In contrast, PCA in brown trout shows weak structuring, with individuals from the Rough River tending to separate from those sampled in the Fiddaunveela and Goulaun rivers (Fig. 4B–C). This pattern is captured by PC1 (4.87% of the variance) and is more clearly expressed along PC3 (4.14%), although substantial overlap between groups remains. Results from the admixture analysis in Atlantic salmon do not reveal any spatial structuring: regardless of the number of genetic clusters (K) retained, individuals assigned to the same cluster are not geographically grouped. The FSTIS estimator identified a single genetic cluster as the best fit for the data (K = 1), while the DLK2 estimator indicated support for six clusters (K = 6). In brown trout, the FSTIS estimator identified a single genetic cluster as the best fit for the data (K = 1), while the DLK2 estimator indicated support for two clusters (K = 2). Two genetic clusters (K = 2) clearly separate fish sampled from the Fiddaunveela and Goulaun rivers from those sampled from the Rough River, with only 6 individuals (out of a total of 295) not matching their geographical genetic cluster (Fig. 4D).

Fig. 4.

A: PCA of Atlantic salmon. B: PCA of brown trout using PC1 and PC2. C: PCA of brown trout using PC1 and PC3. D: individual assignment to genetic clusters inferred by PopCluster (K = 2) in brown trout. Each vertical bar represents an individual, and colours indicate membership probability (Q) to each cluster. For panels B, C and D, trout sampled from the Fiddaunveela and Goulaun rivers were grouped as FIDD_GOUL, while trout sampled from the Rough river were grouped as ROUGH

In salmon, pairwise F_ST values between sites (pairwise distance 1.32 km to 10.33 km) are uniformly low, ranging from 0 to 0.0129 (Table 2). Four out of the ten comparisons are not significantly different from zero according to the bootstrapping CI (Fig. 5A). Furthermore, the Mantel test reveals no significant association between waterway distance and genetic distance (r = 0.24, p = 0.267; Fig. 5A). In trout, similar F_ST values are obtained only between some sites distanced between 300 m and 880 m (Table 3), with a minimum of 0.0004 between Fid_V_01 and Fid_V_02 (630 m apart). Sites above 9 km apart show much higher differentiation than in salmon, with values ranging from 0.028 to 0.045 (2.2 to 3.5 times higher than the maximum pairwise F_ST in salmon). These higher values are between sites on the Rough River and sites on other tributaries, supporting the results from both the admixture analysis and the PCA. Finally, in contrast to salmon, the Mantel test shows a significant correlation between waterway distance and genetic distance (r = 0.96, p = 0.048; Fig. 5B).

Table 2.

Pairwise genetic differentiation (F_ST; upper triangle) and geographic distance (km; lower triangle) among Atlantic salmon sampling sites with ≥25 individuals. Non-significant F_ST values are shown in bold

	GLEN_03	GOUL_04	GOUL_06	MAUM_06	ROUGH_12
GLEN_03	-	0.0100	0.0079	0.0000	0.0029
GOUL_04	10.33	-	0.0044	0.0079	0.0029
GOUL_06	9.01	1.32	-	0.0129	0.0074
MAUM_06	6.91	3.67	2.34	-	0.0012
ROUGH_12	6.11	6.11	4.79	2.69	-

Fig. 5.

Pairwise standardized F_ST values (F_ST/[1 − F_ST]) and their 95% confidence intervals plotted against waterway distance between sampling sites with at least 25 individuals for Atlantic salmon (A) and brown trout (B). Values not significantly different from zero are shown in grey

Table 3.

Pairwise genetic differentiation (F_ST; upper triangle) and geographic distance (km; lower triangle) among brown trout sampling sites with ≥25 individuals. Non-significant F_ST values are shown in bold

	ROUGH_04	ROUGH_05	FID_V_01	FID_V_02	GOUL_01
ROUGH_04	-	0.0050	0.0315	0.0283	0.0393
ROUGH_05	0.88	-	0.0403	0.0384	0.0448
FID_V_01	9.41	10.29	-	0.0004	0.0048
FID_V_02	10.04	10.92	0.63	-	0.0056
GOUL_01	9.11	9.99	0.30	0.93	-

Similar results were obtained for pairwise F_ST, admixture analyses, and PCA when using the subset of the 13 least polymorphic microsatellite markers in trout. PCA showed reduced resolution, with greater overlap among individuals (Supplementary Fig. 3A–B), but retained the same weak clustering pattern, particularly along PC3. Admixture analyses also produced consistent results, with the DLK2 estimator supporting K = 2 and the FSTIS estimator supporting K = 1. At K = 2, individuals from the Fiddaunveela and Goulaun rivers remained clearly separated from those sampled in the Rough River (Supplementary Fig. 3C), although a greater number of individuals were assigned outside their geographical group compared to the full marker set. Similarly, pairwise F_ST values obtained with the 13-marker dataset were very close to those obtained with the full marker set, although slightly lower overall. Importantly, the same pattern of substantially higher F_ST values between sites belonging to the two geographical groups than between sites within the same geographical group was retained (Supplementary Table 4; Supplementary Fig. 2).

Discussion

Contrasting role of breeding dispersal as a driver of gene flow in Atlantic salmon and brown trout

This study presents contrasting patterns of breeding dispersal in two sister species, Atlantic salmon and brown trout, co-occurring within the same catchment. We show that the impact of breeding dispersal on juvenile spatial distribution (I_BD) is 1.77 times higher in Atlantic salmon than in brown trout (Fig. 3; Table 1), with most multiple spawners estimated to move up to 4 km between breeding sites in salmon (with 25% moving beyond this distance), compared with 540 m in trout (with 25% moving beyond this distance). This indicates a markedly greater influence of breeding dispersal on potential inter-tributary gene flow in salmon than in trout. These findings for Atlantic salmon are consistent with those of Taggart et al. [33], who reported that adults spawning at multiple sites commonly disperse over distances exceeding 1 km, with some individuals travelling up to 5.3 km between spawning events. Breeding dispersal in Atlantic salmon can arise from two sources. The first is iteroparity or repeat spawners: individuals that breed in different places across different breeding seasons (here, years), returning to sea between these breeding events [17]. However, their contribution is expected to be low, as repeat spawners typically represent only 3 to 8% of adult spawners [53, 54]. In the Burrishoole, their impact is negligible, as they represent on average less than 1% of the returning salmon (ca. four repeat spawners for 489 returning salmon on average, Marine Institute 2013-2023 [55–64]). Thus, multiple spawning events across different locations within the same breeding season likely constitute most of the breeding dispersal in Atlantic salmon, given that more than half of adult salmon participate in fractional spawning events [33], including 20–40% that spawn more than three times. Consequently, the results of the present study suggest that regular gene flow occurs within distances of approximately 4 km, with reduced but still possible movement between 4 and 12 km within the Burrishoole salmon population. Breeding dispersal can be explained by contrasting behaviours between males and females. Females often build multiple nests (pockets of eggs within redds)—with as many as 14 reported in a season—either within a single redd (the contiguous area of disturbed gravel containing the nests) or distributed among several redds (sometimes up to four) across different stretches or tributaries [65]. Such behaviour likely reduces the risk of catastrophic egg loss. In contrast, males, which do not participate in nest construction, compete for access to spawning females and attempt to participate in multiple fertilization events, with the largest males reported to achieve over 30 fertilizations, whereas the smallest males may achieve three or fewer [65].

In brown trout, the present study identified 6 juveniles whose assigned genetic cluster did not correspond to their geographic sampling group, but instead matched the opposite side of the catchment (>9 km apart; Fig. 4D). Three of these individuals occurred in half-sibling dyads, indicating rare but possible dispersal events over such distances. In the Burrishoole, most brown trout are non-anadromous with only a small number of anadromous trout returning upstream each year (on average 10 adults and 52 finnock, i.e. sea trout that return to freshwater in the same year they migrated to sea and which may or may not subsequently spawn; Marine Institute 2013 [55–64]). Although their contribution is limited, these individuals, together with non-anadromous migrants (i.e., fish migrating exclusively within freshwater throughout their life cycle, see Ferguson et al. [18], may represent spawners capable of travelling large distances. Despite these rare potential long-distance breeding dispersal events, the results of the present study indicate that most adults are unlikely to move more than 1 km within a single breeding season. In addition, habitat preferences may further constrain movement patterns in brown trout. When co-occurring, Atlantic salmon and brown trout favour different spawning grounds reflecting their distinct ecological requirements [66], which could influence the pattern of limited breeding dispersal observed in trout in our study.

Contrasting patterns of genetic structure in sympatric Atlantic salmon and brown trout

We show contrasting patterns of genetic structure in the two sister species. While the salmon population exhibits clear lack of spatial genetic structure, distinct genetic clusters are detected in brown trout, accompanied by substantially higher levels of inter-tributary genetic differentiation.

Genetic homogeneity at the catchment scale in Atlantic salmon

Unlike previous studies that have reported substantial genetic structuring in wild Atlantic salmon populations within large river systems [19–22, 67], the results obtained here do not reveal detectable genetic structure within the much smaller Burrishoole system. The lack of significant genetic differentiation among tributaries indicates extensive gene flow across the catchment. Four out of ten pairwise F_ST values were not significantly different from zero (Table 2), and although the remaining comparisons were statistically significant, their magnitudes were low (mean F_ST = 0.0084) and showed no relationship with spatial distance (Fig. 5). Additionally, both the admixture analysis and the PCA revealed no clustering by sampling location.

These results can be explained by the fact that approximately 90% of Atlantic salmon spawning activity in the Burrishoole River system occurs within a 4 km stretch of the main river [68], representing the largest continuous area of suitable spawning habitat (Fig. 2). Moreover, juvenile salmon sampled in peripheral tributaries of the Burrishoole system, such as Glenamong and Goulaun, despite their apparent geographical isolation, also cluster genetically with the core population, reinforcing evidence of strong connectivity within the catchment. Such connectivity may be partly explained, as discussed earlier, by breeding dispersal. However, it is also likely to be driven by natal dispersal—defined as spawning at a site different from the one where an individual was born. Three forms of natal dispersal can be distinguished.

The first is linked to limitations on the precision of homing, that is, the ability of adults to return to their natal waters to spawn [69]. In small systems such as the Burrishoole, salmon may return to the correct catchment but not necessarily to their exact natal stream, generating regular gene flow among tributaries. This imprecision may partly originate from post-hatching juvenile movements, which limit the scope for fine-scale olfactory imprinting (the process through which salmon memorize the chemical signature of their natal waters) and hence enabling precise homing to natal river habitats. Previous studies have shown that most Atlantic salmon 1+ parr can move 2–3 km from their natal redds, and that fry can move on average ~400 m from their birthplace, with rare upstream movements exceeding 1.2 km and downstream movements exceeding 2 km [70, 71]. Similarly, Kaylor et al. [72] reported that juvenile Chinook salmon typically moved within 1 km upstream and 4 km downstream, with occasional long-distance movements reaching up to 29 km. Such early-life movement could influence homing precision because olfactory imprinting is generally understood to occur during the parr-to-smolt transformation, although Armstrong et al. [73] demonstrated that imprinting can begin earlier in development.

The second form of natal dispersal arises from the movement of precociously mature male salmon parr. In the Burrishoole, 64% of wild male parr were recently found to mature precociously (P. McGinnity, unpublished). These individuals can reproduce through sneak spawning, during which they fertilise eggs laid by females paired with larger anadromous males [17]. They can constitute a major component of the male spawning population [74–78] accounting for 22% to 98% of offspring depending on the population and year, and are estimated to contribute between 20% (Kaufmann, unpublished) and 29% [74] of offspring in the Burrishoole system. Assuming mature male parr disperse at least as much as documented in parr in general, if not more so [79], this behaviour is likely to reinforce gene flow within the spatial scale of parr movement (2–3 km). Importantly, precocious parr reproducing at multiple sites would directly contribute to the breeding dispersal signal quantified in the present study, and therefore represent a component of within-season breeding dispersal. If breeding dispersal were instead interpreted specifically as movement among breeding sites by anadromous individuals only, our estimates would likely be conservative, as precocious parr are expected to disperse over shorter distances.

A third aspect of natal dispersal concerns the contribution of straying individuals originating from other catchments, which can maintain metapopulation dynamics in Atlantic salmon [80, 81]. This process is likely to enhance gene flow at broader spatial scales in salmon compared to trout and may help explain the commonly observed pattern of stronger fine-scale genetic structuring in trout relative to salmon. Moreover, in the Burrishoole system, the very low rate of anadromy in brown trout (see previous section) suggests that such long-distance straying is likely negligible in this species, further reinforcing the contrast between the two.

Fine-scale genetic structure in brown trout

The results obtained for brown trout in this study confirm previous findings indicating the presence of distinct genetic clusters within the Burrishoole catchment [29, 82]. Magee [83] documented significant genetic differentiation between the Goulaun/Fiddaunveela rivers and the lower Rough River. This pattern is confirmed in the present study, which reveals two distinct genetic clusters corresponding to these geographic groups (Fig. 4D; Fig. 5B). Magee also reported a gradual differentiation between lower and upper catchment tributaries, a pattern that we further support through additional analysis (Supplementary Fig. 1). Finlay et al. [29] demonstrated genetic structure at an even finer geographic scale, using a larger sample size and more complete dataset than the present study, between the two closely situated tributaries forming the inflow (Fiddaunveela) and outflow (Goulaun) of Lough Bunaveela. PIT telemetry data from their study showed that although the two trout groups share the same lake-feeding habitat, most individuals return to spawn in their natal stream. Occasional straying was detected—primarily from the inflow to the outflow tributary—but this asymmetric movement did not result in sufficient gene flow to homogenize the populations. Contrasting selective pressures between inflowing and outflowing streams—such as differences in flow direction and hydrodynamic conditions—may contribute to limiting effective gene exchange through divergent local adaptation and different directional migratory responses to water current between inflow and outflow trout [84]. More broadly, brown trout populations living in sympatry within a shared lake habitat but differing genetically, behaviourally, and morphologically have frequently been shown to maintain reproductive isolation [85–87]. Furthermore, brown trout inhabiting different streams or rivers often exhibit marked genetic structure at a range of spatial scales, including scales comparable to those investigated in the present study, due to local adaptation, environmental features or different life-history tactics [30–32].

Limitations

Pairwise F_ST estimates between sampling locations may not be fully comparable between species, as only 1+ individuals were analysed in salmon whereas the trout dataset is dominated by 0+ individuals (285 0+ and 75 1+). Given that older juveniles are expected to have experienced greater post-emergence dispersal, this difference in age composition could lead to lower F_ST values in salmon relative to trout. However, we note that no isolation-by-distance pattern was detected in salmon, and no clustering was observed in either admixture analyses or PCA, regardless of its correspondence with sampling location. This suggests that, while absolute F_ST values may be affected, the overall conclusion of absent genetic structure in salmon within this catchment is robust.

Our estimates of breeding dispersal in brown trout may have been affected by the lack of a gradual increase in distance among sampling sites, in contrast to the more continuous spatial sampling of Atlantic salmon. Differences in site locations and in the number of individuals sampled per site further complicate direct comparisons of intra-dyad distance distributions between the two species, although relative metrics (i.e., I_BD) remain comparable. In addition, the relatively small number of full-sib dyads in the 1+ trout age class introduces increased uncertainty in raw within-full-sib distance estimates for this group. However, as dyads from both 0+ and 1+ individuals are pooled in the estimation of breeding dispersal, this does not bias the overall inference. The approach used in this study to estimate the extent of breeding dispersal assumes that full-sibling families originate from a single breeding event and that the probability of a parental pair spawning at widely separated sites within the same season is negligible. If this assumption is violated, the resulting estimates are likely to be conservative.

Conclusions

This study shows that Atlantic salmon and brown trout, despite being closely related species living in sympatry across the Burrishoole river network (~100 km²), display two contrasting patterns of genetic structure. Salmon appear to engage in behaviours that facilitate substantial gene flow across the catchment, resulting in no detectable population structure even over distances exceeding 12 km. In contrast, brown trout exhibit genetic differentiation at multiple spatial scales, including at fine scales (<1 km), reflecting strong natal tributary fidelity and limited inter-tributary gene flow. One behaviour that partly explains these differences between the two species is breeding dispersal: we show that its extent is much greater in salmon than in trout, with salmon spawners demonstrating the ability to reach distant tributaries between spawning events, whereas most trout remain restricted to movements within their natal tributary. Consequently, when patterns similar to those described here are observed in small catchments, Atlantic salmon populations can be managed as a single unit, without the need for site-specific conservation at the tributary scale, whereas brown trout management must explicitly account for the presence of multiple genetically distinct units, ensuring that conservation actions recognise and preserve this underlying population structure.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: All authors. Methodology development: All authors. Data analysis: T.D. Preparation of figures and tables: T.D. Investigation and data interpretation: All authors. T.D. wrote the main manuscript text. All authors reviewed the manuscript.

Funding

J.K. and P.M. were supported by Science Foundation Ireland, the Marine Institute Ireland and the Department for the Economy, Northern Ireland, under the Investigators Program grant number SFI/15/IA/3028. J.K. and T.D. are currently supported by Taighde Éireann – Research Ireland, under the Pathway programme (grant no. SFI-IRC/21/PATH-S/9671). R.P. and data collection were supported by the Marine Institute, Ireland. T.E.R. is currently supported by Research Ireland (SFI 24/FFP-P/13002).

Data availability

The data and materials supporting this study are available here: https://doi.org/10.5281/zenodo.18552427

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.