Genetic structure and relatedness of brown trout (Salmo trutta) populations in the drainage basin of the Ölfusá river, South-Western Iceland

Sample collection

We sampled 555 brown trout from a total of four lakes and 12 rivers in and around the drainage basin of River Ölfusá, including Lakes Þingvallavatn and Úlfljótsvatn (Fig. 1 and Table 1). Samples of large fish in either lakes or rivers were obtained using nylon gill-nets (25 m × 1.5 m Lundgren series, mesh sizes ranging from 25.0 to 50.0 mm). When used in lakes, the nets were deployed from a boat close to the lake shore where the depth is more than 1.5 m. The nets were set perpendicularly to the coastline and left overnight. Fish were collected after 12 h of fishing, killed, and either processed in situ or taken to the laboratory on ice.

Spawners were sampled in two rivers, Öxará and the lower Ölfusvatnsá. In the latter, nets of mesh size between 60–65 mm were stretched between shores and dragged upstream until up to five individuals were caught. In Öxará, spawners were caught in traps. The fish were sedated with 2-phenoxyethanol (Pounder et al., 2018), and fin clips taken for DNA analysis. The fish were released after having recovered for 30 min in a trap net. Samples of small fish in rivers were taken by electro-fishing. Gear from KC Denmark (Silkeborg, Denmark) was used in the sampling of rivers and streams. It consists of two electrodes (one attached to a 2 m pole) powered by a gasoline engine that produces a current at 200 V AC. The output is run through an AC/DC converter and the resulting current can be applied continuously or pulsed. The stunned fish were collected with a dip net and processed at the sampling location. The fish were immediately sedated upon collection with 2-phenoxyethanol so that a small fin clip could be collected for DNA analysis. The fish were returned to their habitat after they had recovered in a bucket of water with no sedative.

It is not possible to consider individuals from each site as samples from a spawning population since we cannot ascertain that all fish originate from the river or lake where they were caught. However, in the cases where spawning fish were sampled on spawning grounds, or where the habitat is very small and isolated, we assumed that these samples represented proper spawning populations (Table 1).

Sigurður S. Snorrason, lead author on this study has a scientific salmonid fieldwork permit issued by the Directorate of Fisheries, #0460/2021-2.0 and #0042/2014-2.13. All fishing in Lake Þingvallavatn was done with permissions obtained from the farmers in Mjóanes and from the Þingvellir National Park Commission and directors of the park Ólafur Örn Haraldsson and Einar Á. E. Sæmundssen. Fishing in Hestvatn was done with the permit and help of Sigrún Reynisdóttir (landowner at Eyvík farm). Sampling in other rivers was done within the Freshwater Institute monitoring regime of salmonids in Varmá, Sog, Ölfusá, Efra Sog, Villingavatnsá, Ölfusvatnsá, and Leirvogsvatn. The Hengill volcano streams were surveyed with Gísli Már Gíslason (professor emeritus). Samples from Hvítá were provided by Árni Guðmundsson, farmer at Arnarbæli, caught as by-catch in salmon nets.

An anadromous population from Loch Slapin (Isle of Skye, Scotland) was chosen as a reference group in our analyses. The samples from this population were kindly provided by Dr. Colin Adams from the University of Glasgow.

DNA isolation

The DNA extraction was performed using a NucleoMag Tissue purification kit (Macherey-Nagel, Düren, Germany). Fin clips or muscle tissue samples were submerged in a solution containing proteinase K and lysis buffer and digested overnight at 55 °C. Immediately prior to removing the supernatant, a secondary digestion with RNase was performed for 5 min at room temperature. Samples were centrifuged at 5,600 × g for 5 min and the supernatant was retrieved. A suspension of magnetic beads was added, mixed with the sample by pipetting, and allowed to incubate for 5 min at room temperature. After a series of washes, the mixture was incubated in elution buffer (5 mM Tris/HCl, pH 8.5) for 10 min at 55 °C. The supernatant was separated using a magnetic plate and the concentration of DNA was measured via spectrophotometry (Nanodrop). The DNA quality was assessed via gel electrophoresis using 0.6% agarose gels run at 200 V for 45 min. Low quality samples were discarded to avoid preparation of libraries with highly sheared DNA, which would yield a low number of raw reads and possible PCR artifacts (Graham et al., 2015).

Library preparation and sequencing

The methodology is based on Peterson et al. (2012). For every ddRAD library, 96 DNA samples from at least eight different locations were equalized to the same concentration (i.e., 30 or 50 μg μl⁻¹) and distributed in different rows of 96-well plates. Sample DNA was digested with BamHI-HF (final concentration 0.16 U μl⁻¹) in CutSmart buffer (New England Biolabs) for 4 h at 37 °C (final reaction volume 30 μl). After adding NEB 3.1 buffer (New England Biolabs) to the first digestion mixture, a second digestion with ApeKI (final concentration of 0.04 U μl⁻¹) followed immediately for 4 h at 75 °C (final reaction volume 50 μl). Eight adaptors for BamHI-HF cut sites were combined with 12 adaptors for ApeK cut sites to produce 96 unique combinations which were assigned to each sample (Table S1). Adaptors were mixed with samples and T4 ligase (final concentration of 40 U μl⁻¹) and incubated for 12 h at 21 °C. The enzyme was inactivated by incubating the mixture at 65 °C for 10 min.

All 96 samples were pooled together and purified using magnetic beads (Macherey-Nagel, Düren, Germany). Fragments of the appropriate size (i.e., 360–440 bp) were isolated using a 2% agarose Pippin Prep gel cassette and an internal marker for enhanced accuracy (Sage Science, Beverly, MA, USA). A PCR reaction to amplify the resulting fragments was performed using OneTaq reaction mixtures (New England Biolabs) in the following conditions: 72 °C for 3 min, 98 °C for 30 s, 11 cycles of 10 s at 98 °C, 30 s at 65 °C, and 45 s at 72 °C, and a final extension time of 5 min at 72 °C. A further clean-up step was performed with magnetic beads (Macherey-Nagel, Düren, Germany) and the final product was eluted in 5 mM Tris/HCl buffer (pH 8.5) and visualized via electrophoresis in a 2% agarose gel, run for 45 min at 80 V. The final DNA concentration was measured using Qubit fluorometric quantification with the dsDNA BR Assay (Life Technologies, Carlsbad, CA, USA). Library quality was assessed by sequencing on an Illumina MiSeq platform (2 × 150 bp paired-end reads) in four out of the total 13 libraries prepared. Data from MiSeq and HiSeq runs were later combined. Each library was sequenced on a single lane of an Illumina X Ten platform (2 × 150 bp paired-end reads) at the BGI Tech Solutions facility in Hong Kong.

Data analysis

Genetic diversity

A total of 2,597 polymorphic loci from 317 individuals out of 555 originally analysed were retained after the stringent filtering steps outlined in Materials and Methods. These markers were used to address questions about the patterns of genetic differentiation between locations. We estimated population genetic parameters for each sampling site. Heterozygosity varied more than threefold, from 0.080 for Leirvogsvatn (LEI) and Þverá/upper Ölfusvatnsá (THV) to 0.289 for Hvítá (HVI) and Ölfusá (OLF) (Table S4). In most locations the observed heterozygosity was lower than expected, i.e., H_O vs H_E in Table S4 (p < 0.05, Bartlett’s test of homogeneity of variances).

The average nucleotide diversity also varied more than threefold among samples (Fig. 2). The putative anadromous populations and the Scottish reference group (Loch Slapin (SLP), also anadromous) had the highest nucleotide diversity. The individuals from Þingvallavatn (THI) and Úlfljótsvatn (ULF) as well as those from the rivers discharging into them, i.e., Öxará (OXA), lower Ölfusvatnsá (FUS), Villingavatnsá (VIL), and Efra Sog (EFR) showed similar but notably lower levels (i.e., π between 4.30–5.64 × 10⁻⁴), but in Þverá/upper Ölfusvatnsá (THV), isolated from these locations by an impassable waterfall (see Fig. 1), the nucleotide diversity was even lower (π = 3.26 × 10⁻⁴). Similar low genetic diversity was seen in the headwater population of Leirvogsvatn (LEI) (π = 2.58 × 10⁻⁴) (Table S4). Interestingly, sampling sites from tributaries streams of Hengladalsá in three valleys, i.e., from Innstidalur (HIN), Miðdalur (MID), and Fremstidalur (FRE) displayed intermediate levels of diversity, despite being isolated from the lowland locations by waterfalls (Fig. 2). The diversity levels were 2–3 times higher than in the other mountain population i.e., Þverá/upper Ölfusvatnsá (THV). Notably, broader variance in diversity estimates was seen in locations that showed evidence of admixture, i.e., Hestvatn (HES), Hengladalsá at the Miðdalur valley (MID), Ölfusá (OLF), Þingvallavatn (THI) and Úlfljótsvatn (ULF) samples.

Structure and gene flow between brown trout populations

PCA, pairwise F_ST, admixture analyses, and a phylogenetic tree gave congruent results on genetic differentiation. Individuals from many of the sampled locations separated along the first two PC axes (Fig. 3). The first component separated the populations isolated above waterfalls into three distinct clusters, i.e., the Þingvallavatn/Úlfljótsvatn watershed including Þverá/upper Ölfusvatnsá (THV), the southwest Hengill trout, i.e., in tributary streams of Hengladalsá at Innstidalur (HIN), Miðdalur (MID), and Fremstidalur (FRE) valleys, and the headwater local reference population in Leirvogsvatn (LEI). The samples from the putative anadromous populations in the lowlands of Ölfusá (OLF), Hvítá (HVI), and Sog (SOG) as well as those from Hestvatn (HES), a lake connected to Hvítá (HVI), formed the fourth cluster, suggesting a close relation between brown trout from these sampling sites. Interestingly, the fish caught in Varmá (VAR), which we expected to group with the putative anadromous samples because of the connected waterways, clustered with the southwest Hengill populations. Also noteworthy is the fact that the Scottish reference group (SLP) grouped with Ölfusá (OLF) and Varmá (VAR) on the first axis but only separated from the Icelandic trout on the second axis. The third component separated three clusters, one composed of samples from Hestvatn (HES), one with individuals from Ölfusá (OLF), Sog (SOG), Hvítá (HVI), with a few individuals from Hestvatn (HES), and one with the samples from the rest of the locations. The samples from Leirvogsvatn (LEI) separated along the fourth principal component (Fig. S2).

The pairwise estimates of F_ST showed the same patterns and quantified the differences between locations (Table 2). Consistently, the F_ST values among the sample pairs from the Þingvallavatn/Úlfljótsvatn watershed were generally low (F_ST ranging from 0.005–0.134), i.e., group ABOVE (I) in Table 2, except for the isolated Þverá/upper Ölfusvatnsá (THV) population, which had an average F_ST = 0.311. Similarly, fish in Hvítá (HVI), Sog (SOG), and Ölfusá (OLF) (i.e., group BELOW in Table 2) had low pairwise F_ST values (average F_ST = 0.03 when compared among themselves) suggesting that they stem from the same population or belong to sub-populations that have interbred in the recent past. Varmá (VAR) and Hestvatn (HES) are connected to Ölfusá (OLF) and Hvítá (HVI) without any obvious physical barriers to fish migration. Yet the fish from Varmá (VAR) and Hestvatn (HES) deviated strongly from fish caught in Hvítá (HVI), Sog (SOG), and Ölfusá (OLF) (average F_ST = 0.132 and 0.158, respectively).

The highest F_ST values were observed between Leirvogsvatn (LEI) and the rest of the samples from the drainage basin of Ölfusá (i.e., average F_ST of 0.513), which curiously was almost two times higher than the average F_ST (0.298) between the Scottish population (SLP) and the Icelandic samples/populations.

Finally, brown trout from Hengladalsá at Innstidalur (HIN) and Miðdalur (MID) were closely related (i.e., F_ST = 0.054) despite being separated by a waterfall, suggesting downstream gene-flow from the Innstidalur valley (HIN).

The entropy criterion obtained in the admixture analysis reached a minimum at K = 8 (Fig. S3, see Fig. 4 for K = 6–8 and Fig. S4 for K = 9–11). Consistent with other analyses, the Scottish samples formed one cluster and the other Icelandic samples the seven remaining clusters. Brown trout from the tributary streams of Hengladalsá, i.e., at Innstidalur (HIN), Miðdalur (MID), and Fremstidalur (FRE) valleys and downstream in Varmá (VAR) split into two groups (i.e., upper and lower, C1 and C2 respectively), the samples from Hvítá (HVI), Sog (SOG), Ölfusá (OLF), and Hestvatn (HES) were grouped together in C3 while some individuals from Hestvatn (HES) split from the Hvítá (HVI) and Sog (SOG) cluster to form their own group (C4). Most fish caught in the watershed of Þingvallavatn and Úlfljótsvatn clustered together (C5), the exception being the fish from Þverá/upper Ölfusvatnsá (THV) plus a few fish caught in lower Ölfusvatnsá (FUS), Villingavatnsá (VIL) and Þingvallavatn (THI) that clustered separately (C6). Leirvogsvatn (LEI), and the Scottish reference population (SLP) formed distinct clusters, C7 and C8, respectively (Fig. 4).

The admixture proportions varied per location (K = 8, Fig. 5). The results clearly capture the contributions of the small headwater populations, i.e., Innstidalur (HIN), Þverá/upper Ölfusvatnsá (THV), and Hestvatn (HES) into downstream locations. The data do not indicate downstream flow of genes from the Þingvallavatn/Úlfljótsvatn system into the putative anadromous populations in the lowlands.

A unique genetic component from Þverá/upper Ölfusvatnsá (THV) appeared downstream in the fish caught in the lower Ölfusvatnsá (FUS) (directly below a waterfall) and Villingavatnsá (VIL). The latter river runs parallel to Ölfusvatnsá in the lower reaches and has its mouth in Þingvallavatn ca. 700 m to the south of the former (Fig. 1). This genetic component was present in most of the fish caught in Þingvallavatn (THI) and to a lesser extent, in the fish from Úlfljótsvatn (ULF) (Fig. 5), but notably not in fish caught spawning in Öxará (OXA).

There was clear evidence of downstream gene flow from the populations inhabiting the streams on the south side of the Hengill volcano. Again, Hengladalsá at Innstidalur (HIN) and Miðdalur (MID) were closely related (i.e., F_ST = 0.054) despite being separated by a waterfall, and the ancestry plots suggested a downstream flow of genetic material from Innstidalur (HIN). The samples from the upper parts of Hengladalsá (i.e., at the Innstidalur valley, HIN) showed a unique component that progressively decreased downstream. The samples from Varmá (VAR) clustered with individuals from Hengladalsá (Fig. 3) and the ancestry plot showed a high proportion of a genetic component from Hengladalsá (Fig. 4). This component also appeared in putative anadromous fish in the lowlands, i.e., Hvítá (HVI), Sog (SOG), Ölfusá (OLF) (Fig. 4). Note also that the fish caught in Varmá (VAR) have a significant genetic contribution from Hvítá (HVI)/Sog (SOG)/Ölfusá (OLF) (Table S5), reflecting the lack of barriers to upward migration from Ölfusá (OLF).

Mixing or contribution of genes between Hestvatn (HES) and the anadromous fish in Hvítá (HVI), Sog (SOG), and Ölfusá (OLF) was also evident, but the data also indicate that some Hestvatn (HES) trout belong to a resident population. A genetic component from the lowlands, which was most predominant in the Sog (SOG) samples, appeared in some individuals caught in Hestvatn (HES).

A phylogenetic analysis with RAxML while largely congruent with the PCA revealed several extra observations (Fig. 6). The samples from the Þingvallavatn/Úlfljótsvatn watershed, i.e., Þingvallavatn (THI), Öxará (OXA), lower Ölfusvatnsá (FUS), Þverá/upper Ölfusvatnsá (THV), Villingavatnsá (VIL), Efra Sog (EFR), Úlfljótsvatn (ULF) and those from the south-western slopes of Hengill, i.e., Varmá (VAR), Fremstidalur (FRE), Miðdalur (MID), Innstidalur (HIN) formed two distinct clades separated by a large phylogenetic distance. It should be noted that the brown trout in the southwestern slope of Hengill and the river Þverá are relatively close geographically (ca. 1 km apart, separated by a mountain ridge) but the streams drain in opposite directions.

The headwater population of Leirvogsvatn (LEI) and the reference group from Scotland (SLP) formed their own distinct clades (Fig. 6). A well-supported branch containing Leirvogsvatn (LEI) was placed near the Þingvallavatn/Úlfljótsvatn branch (Fig. 1).

The clade distribution of individuals from the lowland putative anadromous populations was less clear, with individuals from Sog (SOG), Ölfusá (OLF), and Hvítá (HVI) landing on different poorly separated branches. Samples from Hestvatn (HES) grouped in two clusters on the same well supported branch. The samples from Varmá (VAR) connected to the stem of the well supported Hvítá (HVI)/Sog (SOG)/Ölfusá (OLF) branch. Samples upstream of Varmá (VAR) formed a well-supported clade containing a mix of individuals from the lower parts of Hengladalsá (i.e., Miðdalur (MID) and Fremstidalur (FRE) valleys) and a sub-branch with individuals from the upper part of the river, i.e., Innstidalur (HIN) as well as some from Miðdalur (MID).

Effective population size and test for a population bottleneck in Þingvallavatn

We wanted to estimate the effective size of the populations sampled at each location. The filtering approach designed to retain more individuals and markers yielded 377 individuals and 5,353 polymorphic loci, but the LD correction reduced this to 3,505 markers (Table S6) All sampled locations showed low effective population sizes, with the smallest value in Leirvogsvatn (LEI) (i.e., N_E = 3.4) (Table 3). Brown trout from Sog (SOG), presumably anadromous fish, had the highest effective population size (i.e., N_E = 71.2). Notably, the estimate for Öxará (OXA) was the second largest (i.e., N_E = 56.5) while that of lower Ölfusvatnsá (FUS), showed an N_E = 14.9 (Table 3). Recall, both drain into Þingvallavatn and clustered together by genetic relatedness.

To test for a bottleneck in the population of Þingvallavatn, we used a different dataset (i.e., 3,019 polymorphic loci, 373 individuals) using haplotypes and excluding sampling locations with <20 individuals. For reference, other locations with n > 20, including the Scottish reference group, were also tested.

None of the three runs for the three different models (i.e., IAM, SMM, TPM) yielded significant results for the Wilcoxon tests for heterozygosity excess, except for the reference group SLP with the IAM model. All other sampling locations showed significant heterozygote deficiency (p < 0.001).

Structure and gene flow between brown trout populations

The results (F_ST values, PCA and the phylogenetic tree) showed a pattern and degree of genetic structure that fit expectations based on the present topography and connectedness of waterways in the western part of the Ölfusá watershed. The headwater populations, which are located above impassable waterfalls, showed clear signs of isolation from the putative anadromous lowland populations, which may exchange genetic material more freely. Importantly, individuals from Úlfljótsvatn, Þingvallavatn, and its tributaries, i.e., Úlfljótsvatn (ULF), Þingvallavatn (THI), Öxará (OXA), Ölfusvatnsá (FUS), Villingavatnsá (VIL), and Efra Sog (EFR) clustered together and most of their pairwise F_ST values were low (range 0.005–0.134), implying that the populations of these lakes and tributaries are genetically very similar. As was expected, the Þverá/upper Ölfusvatnsá (THV) population, located above an impassable waterfall, was the only one forming a genetically distinct sub-group within this clade. The significant ancestry component from this population in lower Ölfusvatnsá (FUS), Villingavatnsá (VIL), and Þingvallavatn (THI), suggests significant downstream gene flow from Þverá/upper Ölfusvatnsá (THV). The observed pattern of admixture may stem from remnants of spawning populations in lower Ölfusvatnsá and Villingavatnsá, which, due to geographic proximity, would be affected by down-stream gene flow from the Þverá/upper Ölfusvatnsá (THV) population, an influence not seen in the Öxará spawners.

Although written records of brown trout spawning sites are scant, it is generally thought that prior to the construction of the dam the trout of Þingvallavatn originated from several sub-populations of spawners. The best-known spawning locations are in the rivers Öxará, lower Ölfusvatnsá, Villingavatnsá, and close to and in the outlet at Efra Sog (Jóhannsson & Jónsson, 2002) and according to farmers around the lake, trout was also known to spawn at several cold-water spring sites within the lake (Skarphéðinsson, 1996). The lack of a subgroup structure reflected by the low F_ST values among the fish caught in these locations and the dominant admixture component from the Öxará population (OXA) seen at present may to a large extent be a consequence of the efforts to reinvigorate the trout of Þingvallavatn.

It is notable that the parr caught in Villingavatnsá (VIL), which had no releases of fingerlings from Öxará (OXA), were the most genetically distinct among the samples from Þingvallavatn, Úlfljótsvatn and its tributaries (F_ST values ranging from 0.058–0.134). Moreover, they show the most definite genetic component from the isolated population in Þverá/upper Ölfusvatnsá (THV). The fish caught in lower Ölfusvatnsá (FUS) were more mixed in this respect, some showing a strong component from Þverá/upper-Ölfusvatnsá (THV) while others mostly had the Öxará (OXA) component. This suggests that the releases of parr and planting of eggs from the Öxará spawning population lead to mixing of these components in the trout presently spawning in lower Ölfusvatnsá (FUS). The planting of eggs and releasing of juveniles may have been crucial, as the hatched fish would have experienced the imprinting of the natal river conditions. As the Þverá/upper Ölfusvatnsá (THV) population consists of small trout, but the lake fish can reach very large sizes, it remains an open question whether this gene flow influences traits in individuals of mixed ancestry. When some Irish lakes were stocked with adults for a period of 7 years, the genetic material of the native populations remained intact because the introduced fish did not spawn in the streams, as they probably lacked the imprinting of the conditions of the natal streams at an early age (O’Grady, 1984).

Similar patterns of phylogenetic separation, with directional gene flow, were seen in mountain tributaries of Hengladalsá. These trout clustered on a well-supported branch (98%) of the phylogenetic tree. Although we only had three samples from the upper Hengladalsá (i.e., from the Innstidalur valley (HIN), isolated above impassable waterfalls), the data suggest that these isolated tributaries harbour a genetically distinct population. Below these waterfalls, Miðdalur (MID) and Fremstidalur (FRE) had 15/40 and 1/15 fish with an Innstidalur (HIN) ancestry proportion >0.5, respectively. Most fish caught in the tributary streams of Miðdalur (MID) (25/40) and Fremstidalur (FRE) (14/15) showed a different genetic component that, somewhat unexpectedly, was found dominant in the fish from Varmá (VAR). Notably, this component was also found in trout from the Hvítá (HVI)/Sog (SOG)/Ölfusá (OLF) system. As might be expected, all the fish from Varmá (VAR) had a small but significant ancestry proportion (0.08–0.25) from the Hvítá (HVI)/Sog (SOG)/Ölfusá (OLF) system. The reasons for the shift in genetic composition from Innstidalur (HIN) to Miðdalur (MID) are unclear but we hypothesize that this may be caused by differences in habitat. The Innstidalur (HIN) habitat only consists of cold springs whereas the fish from the Miðdalur (MID) and Fremstidalur (FRE) valleys were caught in streams influenced by geothermal activity which drastically changes the annual temperature regime and ecological cycles of these streams (O’Gorman et al., 2016; Ólafsson, 2019). It seems plausible that adaptation to this environment may have led to genetic differentiation of these populations. This could also explain the dominance of the Miðdalur (MID) genetic component in the fish from Varmá (VAR) as this river is also significantly influenced by geothermal springs.

Considering the historically large trout populations of the Þingvallavatn/Úlfljótsvatn system (Skarphéðinsson, 1996) it is somewhat surprising not to see some genetic contributions from these populations in the putative anadromous populations downstream of the Írafoss and Kistufoss waterfalls. While it seems inevitable that brown trout would accidentally be swept downstream of the waterfalls, such escapees do not mix effectively with downstream populations.

It is notable that fish from Leirvogsvatn (LEI), despite having the highest F_ST values with respect to the samples from the Þingvallavatn/Úlfljótsvatn watershed (i.e., OXA, THI, ULF, FUS, EFR, THV), are close to the Þingvallavatn/Úlfljótsvatn fish on the first PCA axis and are placed together, on a well-supported branch, in the phylogenetic tree. This suggests a common origin and early split between these populations. Presumably the early isolation of trout in the Þingvallavatn/Úlfljótsvatn system resulted in strong selection against downstream migration and rapid adaptation to a resident lifestyle, a process that would have been greatly enhanced by high productivity of benthic invertebrate prey and opportunities for switching to piscivory in these lakes. Migratory behaviour is a quantitative trait with a genetic component (Palm & Ryman, 1999; Kendall et al., 2015). A highly productive freshwater environment has also been linked to a reduced likelihood of anadromy (Finstad & Hein, 2012). Similarly, our data indicate that the early establishment of residency of the brown trout in Þingvallavatn and Úlfljótsvatn has contributed significantly to the high degree of genetic differentiation observed between them and the populations of the Hvítá/Sog/Ölfusá system.

Although fish from the connected lowland waterways in Hestvatn (HES), Hvítá (HVI), Sog (SOG), and Ölfusá (OLF) scattered on the same branch of the tree, most individuals caught in the same water body tended to cluster together on well supported sub-branches. In Hestvatn (HES), most of the fish clustered on a well-supported (100%) sub-branch while a few fish had a high proportion (around 50%) of a genetic component from the Hvítá (HVI)/Sog (SOG)/Ölfusá (OLF) system. Some individuals caught in Hestvatn had ca. 30% of the genetic component from the lake and about 50% of the Hvítá/Sog/Ölfusá component (Table S5), indicating genetic admixture. This suggests that Hestvatn harbours two sub-populations, a distinct resident population and migrants from River Hvítá or trout from interbreeding between those stocks. The stream connecting Hestvatn to Hvítá is less than 2 km long and there is no impassable obstacle that could prevent migration under normal circumstances (M. Jóhannsson, 2021, personal communication). During floods in Hvítá glacial water flushes into Hestvatn. According to a farmer by the lake, such floods can bring spurts of migrating Atlantic salmon, Arctic charr and brown trout into Hestvatn (S. Reynisdóttir, 2021, personal communication).

Small effective population size, and low genetic diversity

The N_E estimates for several locations in the watershed of Ölfusá showed small values, the exception being the spawning trout from Öxará (OXA), and the fish sampled in Sog (SOG) which had larger N_E, and the population in Leirvogsvatn (LEI) with a very small N_E.

The observed variation in N_E values may reflect habitat constrains that limit the number of sexually mature individuals coming to the spawning grounds.

Although we have no knowledge of spawners or spawning locations in the mountain streams of Hengill, it seems plausible that the adult population size could be limited by the availability of food in the small habitats of the mountain streams flowing into Hengladalsá (i.e., HIN, MID, FRE) and Þverá/upper Ölfusvatnsá (THV), resulting in the small N_E values observed. This is in stark contrast to the situation in Þingvallavatn, which offers vast habitats and favourable conditions for growing immature fish. Presently, several thousand individuals enter Öxará each year to spawn (ca. 3,000 in 2019, see Laxfiskar, 2022). Although the N_E observed in this population (OXA) was almost four times larger than that of Þverá/upper Ölfusvatnsá (THV), it seems low considering the number of spawners. This strongly suggests a different type of habitat constraint, i.e., that N_E is limited by the size of the spawning area. The number and quality of suitable spawning areas can have a direct impact on the effective population size (Heggenes et al., 2009). For females, the competition for the chance to reproduce is associated with an increased risk of nest overlapping and embryo mortality because of delayed spawning (Belmar-Lucero et al., 2012). It seems most likely that the disparity between the N_E estimate of the Öxará population and the number of spawners results from fierce competition and nest overlap in a limited spawning area, a situation that results in an outcome where only a small proportion of spawners are successful in leaving viable offspring.

Long isolated populations in small habitats with low carrying capacity are prone to have low effective population sizes (Funk et al., 2016). The small N_E values observed in our study are in line with the literature. Small isolated brown trout populations showed N_E < 100 in central Italy streams (Splendiani et al., 2019; Rossi et al., 2022), northern Spain rivers (González-Ferreras et al., 2022), one oligotrophic lake in Sweden (Charlier, Laikre & Ryman, 2012), and in a metapopulation of a Swedish lake system (Andersson et al., 2022). Linløkken, Johansen & Wilson (2014) found N_E values between 9.2–1,222.3 that correlated positively with size habitat in brown trout populations of a lake watershed in south-east Norway.

It is important to consider that the linkage disequilibrium method used to estimate N_E may not only reflect the N_E per generation, but also the number of breeding individuals per year (Nb), or something in between. For a species like brown trout (with overlapping generations and repeated spawning), Nb is expected to be lower than N_E (Waples et al., 2013). In general, Nb in brown trout populations has been shown to be significantly lower than N_E (Charlier, Laikre & Ryman, 2012; Serbezov et al., 2012). Furthermore, a downward deviation of the N_E estimation may be caused by genetic admixture (Waples, 2006), the presence of siblings (Whiteley et al., 2013; Linløkken, Johansen & Wilson, 2014), or loci under selection (Linløkken, Johansen & Wilson, 2014), while the presence of immigrants could cause an upward bias (Serbezov et al., 2012). Although we attempted to correct for some of these factors, the possibility of a downward bias remains.

The small effective population size values observed in this study may pose a threat to some of these populations. An effective population size in the range of 1,000–5,000 seems to be an adequate number to maintain the genetic diversity of populations in the long term (Lynch & Lande, 1998). It has been proposed that N_E of 500 may be enough (Frankham, 1995; Andersson et al., 2022), which is a value far above all the estimates in this study. Populations with small effective population size have a reduced genetic diversity that compromises their adaptability (Frankham, 1996).

Our results revealed a clear pattern of genetic diversity by geography and waterway topography. The small headwater populations of Þverá/upper Ölfusvatnsá (THV) and Leirvogsvatn (LEI) showed the lowest diversity of the populations studied, with the average nucleotide diversity being ca. 70% lower than that of the putative anadromous lowland populations. The nucleotide diversity was significantly higher in samples from Úlfljótsvatn (ULF) and Þingvallavatn (THI) and the accessible part of its tributary streams but ca. 50% lower than that of the fish caught in the lowland rivers below the Írafoss and Kistufoss waterfalls.

The low values of observed heterozygosity (H_O) and nucleotide diversity estimates (π) in most of the headwater locations sampled are probably the result of many generations in isolation. However, in the case of Leirvogsvatn (LEI) and Þverá/upper Ölfusvatnsá (THV), habitat constraints could also play a role, e.g., restricted spawning habitats (in Leirvogsvatn, LEI) and restricted winter habitat and adaptation to the ecological effects of the thermal springs (in Þverá/upper Ölfusvatnsá, THV). Considering the low levels of genetic diversity in Leirvogsvatn (LEI), Þverá/upper Ölfusvatnsá (THV), and the populations of Úlfljótsvatn (ULF) and Þingvallavatn (THI) and its tributaries, the relatively higher genetic diversity in the mountain streams in Hengladalsá is odd. This may reflect admixture caused by recent stocking or population substructure. The six fish from Innstidalur (HIN) were sampled in cold springs but the fish from Miðdalur (MID) and Fremstidalur (FRE) were sampled in several very small, thermally influenced streams (O’Gorman et al., 2016; Ólafsson, 2019) that enter the much colder Hengladalsá separately. Pooling small samples from these thermally diverse habitats possibly masks a hidden population structure thus inflating estimates of genetic diversity. An alternative, but not exclusive scenario, is the possibility of active releases of brown trout from lowland populations into Hengladalsá. Guðjónsson, Guðmundsson & Jónsson (1983) reported the release of brown trout in Varmá and/or Hengladalsá in the early 1960’s. These records are unclear, however, as there is no indication whether this was done below and/or above the impassable waterfalls in Hengladalsá. Another factor to consider when analysing the values of H_O and π is that the removal of loci that deviate from Hardy-Weinberg equilibrium, as was done in this study, is known to reduce H_O and π estimates (Shafer et al., 2017). However, since this was applied equally to all populations, the relative differences in genetic diversity should have been preserved.

Did the dam and restoration efforts influence Lake Þingvallavatn brown trout?

Despite the observed drop in catches of trout in Þingvallavatn after the dam and power plant construction (Malmquist, Snorrason & Skúlason, 1985) and a sustained period (1970–1990) of very low catches and poor recruitment, our data showed no signs of a recent population bottleneck in the Þingvallavatn/Úlfljótsvatn watershed. While the severe drop in catches of brown trout in most areas of Þingvallavatn in the 1960’s and 1970’s may have been mostly caused by the loss of the largest spawning ground during the dam construction, it is difficult to explain how this could have caused the prolonged lows in recruitment at other spawning grounds. The history of trout fishing in Þingvallavatn before the dam suggests sub-populations of trout in Þingvallavatn may have fluctuated considerably over time, in some cases due to heavy fishing (Skarphéðinsson, 1996). Long-time (decadal) fluctuations in temperature and a steady global warming trend may also have played a role in these fluctuations. In northern latitudes global warming appears to favour brown trout relative to Arctic charr (Svenning et al., 2022) and this may explain the explosive growth of the trout population in Þingvallavatn in the last two decades (Q. J. B. Horta-Lacueva, F. Finn, F. Ingimarsson, H. R. Ingvarson, H. Xiao, K. H. Kapralova, L. Ponsioen, M. Lagunas, M. De La Camara, N. Eskafi, S. M. Stefánsson, S. S. Snorrason, 2022, unpublished data).

The stocking efforts in Þingvallavatn seem to have been successful despite some differences in the physical characteristics of rivers where the eggs originated, i.e., Öxará, and where they were planted, i.e., in the lower Ölfusvatnsá and near the outlet in Efra Sog (Jóhannsson & Jónsson, 2000a; Jóhannsson & Jónsson, 2016b; Q. J. B. Horta-Lacueva, F. Finn, F. Ingimarsson, H. R. Ingvarson, H. Xiao, K. H. Kapralova, L. Ponsioen, M. Lagunas, M. De La Camara, N. Eskafi, S. M. Stefánsson, S. S. Snorrason, 2022, unpublished data). Recruitment of juvenile trout from spawners in the lower Ölfusvatnsá seems plentiful at present (Q. J. B. Horta-Lacueva, F. Finn, F. Ingimarsson, H. R. Ingvarson, H. Xiao, K. H. Kapralova, L. Ponsioen, M. Lagunas, M. De La Camara, N. Eskafi, S. M. Stefánsson, S. S. Snorrason, 2022, unpublished data). There can be little doubt that the planting of embryos and release of juveniles derived from the Öxará spawners has affected the genetic structure of the meta-population of trout in the Þingvallavatn/Úlfljótsvatn system. This is reflected in the strong genetic component from the Öxará spawners seen in samples from all other locations.