Genetic mixed-stock analyses of Baltic salmon – a review

Stock proportions in commercial salmon catches from the Baltic Sea have been analysed as a part of annual WGBAST reports since 2005. Stock proportions have been analysed yearly for Finnish catches from the Gulf of Bothnia. Swedish catches from the Gulf of Bothnia were also analysed yearly as a part of the WGBAST reports until 2017 (ICES, 2017a). Stock proportions in Swedish coastal salmon catches have further been analysed for national reports in 2014 and 2015 (Öster-gren et al., 2014; 2015). Finnish and Estonian commercial catches from the Gulf of Finland were most recently analysed in 2019 (ICES 2019). Stock proportions in the commercial catches from the Main Basin have been analysed for the 2015 and 2017 WGBAST reports (ICES, 2015; 2017a).

In addition, the presence of salmon from individual stocks in either commercial or experimental catches from the Baltic Sea have been analysed and presented in several peer-reviewed publica-tions (Koljonen and McKinnell, 1996; Koljonen and Pella, 1997; Pella and Masuda, 2001; Koljonen et al., 2005; Koljonen, 2006; Palm et al., 2008; Vuori et al., 2012; Whitlock et al., 2018).

The estimates of stock proportions in commercial catches of the Baltic Sea as reported in the annual WGBAST reports (ICES, 2005–2020) are based on stock assignment with DNA-ellite data. In short, each sampled fish from commercial catches is genotyped with 17 microsat-ellite loci and the smolt age of each sampled fish is determined. The genotypes and smolt age are then compared to a baseline dataset, which consists of genotypes and smolt age distributions of different salmon stocks from different spawning rivers around the Baltic Sea. Based on the com-parison of genotypes and smolt age with the baseline data, each salmon from commercial catches is then assigned the most probable stock of origin. Presently, the baseline dataset consists of 4453 individual salmon from 39 individual stocks from six different countries, each genotyped with 17 microsatellite loci. The stock assignment is based on Bayesian inference as implemented in the software BAYES (Pella and Masuda, 2001), which allows integrating smolt age information into the stock assignment, making especially the distinction between wild and hatchery origin stocks more reliable (Koljonen, 2006).

In general, the reliability of stock assignment with 17 microsatellite markers combined with smolt-age data is quite high. For example, the mean proportion of times an individual was as-signed to a particular stock in 1 000 iterations of MCMC in the 2017 Gulf of Finland catch samples was 0.95 (N = 411 individual salmon). In the Bothnian Bay catch samples from 2020, the mean proportion of assignment to a certain stock was slightly lower, 0.88 (N = 111 individual salmon, regular fishing season), due to the genetic similarity of wild stocks from the Kalixälven and Tor-nionjoki rivers (see Miettinen et al., 2021) as well as the mixed genetic background of the river IIjoki hatchery stock (Säisä et al., 2003). It must also be noted that stock proportion estimates prior to 2008 were based on eight microsatellite markers, which is likely to reduce their power of res-olution, especially of genetically similar stocks.

The most comprehensive data on stock proportions in commercial catches are from the Gulf of Bothnia, where the stock proportions in Finnish catches from three fishing areas along the coast have been estimated annually for the past 20 years. In summary, the data show that the great majority of salmon caught by the Finnish commercial fisheries originate from wild stocks of riv-ers Tornionjoki and Kalixälven, and hatchery stocks with their genetic origin in the river Torni-onjoki (means in the regular fishing season 2009-2020: 68% wild stocks and 29% hatchery stocks).

In addition, a small proportion of the catches originate in Swedish hatchery stocks (mean 2009–

2020: 2%). Salmon originating in Swedish wild stocks other than river Kalixälven are only caught occasionally; their total proportion of the Finnish salmon catch in the Gulf of Bothnia always has been <1%. No salmon from AUs 4–6 have been encountered in the commercial catches from the Gulf of Bothnia. The proportion of wild stock salmon from the rivers Tornionjoki and Kalixälven has increased following the change in temporal fishing regulations implemented in Finland since 2017, which has allowed an earlier start of fishing with limited number of fykenets (mean pro-portion of wild stock salmon in the advanced fishing season catches 2017–2020: 78%). There is, however, annual variation in the proportions of wild and hatchery origin stocks in the catches (proportions in regular season catches from 2009–2020: wild origin 58%–82%, hatchery origin 18%–38%).

Stock proportions in Swedish commercial catches from the Gulf of Bothnia were last reported in the 2017 WGBAST report (ICES, 2017a). As for the Finnish Gulf of Bothnia catches, a large pro-portion of the caught salmon originated in the wild northernmost Bothnian Bay rivers Torni-onjoki and Kalixälven (63% in 2013–2016 Finnish catches, 38% in 2013–2016 Swedish catches).

Compared to the Finnish catches, there were less salmon from the Finnish hatchery stocks (rivers Kemijoki, Simojoki, Iijoki, and Oulujoki). The other main difference was that in these Swedish catches, there were salmon from additional wild Swedish rivers, especially Byskeälven and Vindelälven (mean proportions in 2013–2016: 12% and 10%, respectively). In total, the proportion of wild salmon has been higher in the analysed Swedish catches from the Gulf of Bothnia (mean 2009–2016: 82%) than in the Finnish catches (mean 2009–2016: 70%) (ICES, 2017a). It must be

noted, however, that the Swedish catch data analysed until 2017 was just collected from a limited number of fishing sites, and that the stock composition estimates from the western Gulf of Both-nia are much more influenced by the geographic position than on the eastern side along the Finnish coast, where a more homogenous stock mixture is harvested (ICES, 2017a; below).

Salmon from the weakest Swedish stocks have only appeared once in Swedish commercial catches in 2006–2016 in the Gulf of Bothnia (Ljungan: 1% in 2010). The only stock from AUs 4–6 that has been caught by the Swedish fishery in 2006–2016 in the Gulf of Bothnia has been the Finnish hatchery stock genetically originating in the river Neva (2% in 2007).

Stock proportions in the Swedish coastal fishery in 2013–2014 have been reported in a national report by Östergren et al. (2015). Compared to the above ICES-analyses, a significantly larger number of fishing sites distributed along the Swedish coastline were included (18 sites, whereof eight were sampled in both 2013 and 2014, comprising a total of 2 850 individuals; Östergren et al., 2015). The same genetic data from 2014 have also been analysed by Whitlock et al. (2018; in press) in peer-reviewed methodological studies on the integration of genetic analysis of mixed stocks with a population dynamics model. In these analyses, samples from the Finnish coastal fishery in the same fishing year were also included. The genetic data from 2013 and 2014 (in combination with catch data from 2019) have also been analysed, using the same model, in a national report focusing on stock composition in Swedish coastal catches (Dannewitz et al., 2020b, see Section 4.5.3.2).

The main difference between the stock abundance estimates in catches presented by Östergren et al. (2015), Whitlock et al. (2018) and Dannewitz et al. (2020b) compared to those for Finnish and Swedish sea catches reported in earlier WGBAST reports was that the coastal Swedish catches were mainly composed of salmon from the rivers (wild or reared) closest to the catch sites. Also, the river Kalixälven and Tornionjoki wild stocks, which dominate the Finnish and Swedish catches reported by WGBAST, were nearly absent in the Swedish coastal catches (other than in the catches from near their own river mouths). Whitlock et al. (2018) further showed that the migration patterns can vary greatly among different Baltic Sea salmon stocks, meaning that there may exist strong variation in stock compositions at different times in a given area. Combining genetic marker data with information on population dynamics and movement provides a means for temporal and spatial regulation of fishing efforts to target reared and healthy wild stocks while avoiding weak ones (Whitlock et al., 2018). See Section 4.7.1 for a discussion on future potentials of including results from the Whitlock et al. (2018) coastal model as prior information into the regular WGBAST stock assessment.

Estimates of stock proportions in Finnish commercial catches from the Åland Sea are available from 2000–2016 (ICES, 2017a). Again, the largest proportions in the catches are from the rivers Kalixälven and Tornionjoki wild stocks (means 2000–2016: 23% and 34%, respectively). There are also small, but yearly varying proportions of salmon from the northern Bothnian Bay stocks as well as from the stocks from Swedish rivers (all means 2000–2016: 0–6%).

Stock proportion estimates from catches in the Gulf of Finland have been last reported in the 2019 WGBAST report (ICES, 2019) and include Estonian (2016–2018) and Finnish (2009–2018) coastal fisheries. The stock with the largest proportion in the Estonian catches has been the river Kunda stock, which includes both wild and hatchery origin fish (mean 2016–2018: 40%). The catches have also included salmon from rivers Keila (wild origin, mean 2016–2018: 17%) and Narva (hatchery origin, mean 2016–2018: 12%). The Estonian catches have further included salmon from the wild and hatchery stocks from AU5: Salaca (wild origin, mean 2016–2018: 3%) and Daugava (hatchery origin, mean 2016–2018: 6%). The Finnish coastal catches from the Gulf of Finland are mainly comprised of stocked salmon of river Neva origin (mean 2009–2018: 49%) and salmon from the northern Bothnian Bay rivers (Tornionjoki, Kalixälven, Oulujoki, mean 2009–2018: 43%). Salmon from other Swedish rivers have only been found occasionally in these

Finnish catches (<1%), and salmon from AU5 and AU6 rivers only have been found in the 2018 Finnish catch: Daugava, AU5, hatchery origin: 3% (95% Confidence bounds: 1–5%); Keila, AU6, wild origin: 1% (95% Cb: 0–2%).

Stock proportion estimates for the Main Basin salmon fishery include data from commercial catches in selected years from Danish (2006, 2010–2016), Finnish (2006–2007, 2009–2012), Latvian (2006), Polish (2006–2016), and Swedish (2006–2007, 2010–2012) offshore catches. Catch data from different countries have been pooled and analysed together for each year. The stocks with largest proportions in the Main Basin salmon catches have been the wild stocks from rivers Kalixälven and Tornionjoki (means 2006–2016: 14% and 37%, respectively) (ICES, 2017a). The rest of the catches have been mainly composed of salmon from Swedish wild and reared river stocks (mean proportions for each stock in 2006–2016: 1–6%). The catches from the Main Basin have also in-cluded a few salmon from the AU5 hatchery stocks of rivers Gauja, Daugava, and Neumunas (2006–2016 means: 0–1%; 95% confidence bounds always including zero). Salmon from the wild Salaca (AU5) and mixed Luga (AU6) stocks have also been found in low proportions (1–2%) in the Main Basin catches from 2006–2007 and 2010–2011 (ICES, 2017a).

Stock proportion estimates for different areas of the Baltic Sea have also been published in peer-reviewed publications by Koljonen (2006), Palm et al., (2008), Vuori et al. (2012), and Whitlock et al. (2018). Koljonen (2006) analysed stock proportions in catches from the Gulf of Bothnia, Gulf of Finland, Åland Sea, and the Main Basin with 8 microsatellite markers, and with a baseline of 32 stocks. The data are in large part the same that have been published in the WGBAST reports in 2005 and 2006 (ICES, 2005–2006). The catch data in the WGBAST reports are based on sam-pling of commercial catches in open sea fisheries in winter and early spring in the Main Basin and coastal fisheries in summer in other areas. Palm et al. (2008) used a slightly extended baseline data from Koljonen et al. (2008) to estimate stock proportions in catches taken in late autumn in 2002–2003 from the Main Basin. Their stock proportion estimates were very similar to those in Koljonen et al. (2006): most of the salmon were from the Bothnian Bay stocks. There were no salmon from the currently weak stocks of river Ljungan or Emån. The only AU 4–6 stocks that were present in the catches were the hatchery stocks of rivers Neva (stocked into Finnish rivers;

1%; PI: 0–3%) and Gauja (1%: PI: 0–3%).

Vuori et al. (2012) have in addition analysed stock proportions of salmon catches in late autumn and winter from the Bothnian Sea, Gulf of Finland, and the Main Basin. The catch data analysed by Vuori et al. (2012) are a combination of commercial catches and samples collected for scientific purposes in 2006–2007. The data in Vuori et al. (2012) provide information on migration patterns of salmon from Baltic Sea stocks, but does not reflect the proportions of different stocks in today’s commercial catches as there is no longer commercial fishing of salmon in autumn and winter in the middle and northern Main Basin and northern Baltic Sea. The largest numbers of salmon in the samples from the Main Basin were from the river Tornionjoki wild and hatchery stocks (63/141 salmon, 45%). The number of samples from the Bothnian Sea was low (n = 31) and in-cluded individual fish (1–6) from wild and hatchery stocks of Finnish and Swedish rivers, and a few fish from the mixed stock of the Russian river Luga (n = 2) and hatchery stocks of the Latvian rivers Gauja and Daugava (n = 1 and n = 3, respectively). The most numerous stocks in the sam-ples from the Gulf of Finland were the Finnish river Neva hatchery stock (30/55 salmon) and the mixed stock from the river Luga (15/55 salmon). The rest of the salmon in the Gulf of Finland catch samples were from hatchery stocks of rivers Narva (n = 2), Gauja (n = 1), and Daugava (n = 7).

Whitlock et al. (2018) estimated stock proportions in catches from 18 coastal fishing sites in Fin-land and Sweden. Their findings also followed those reported in the annual WGBAST reports:

stock composition along the Finnish coastline was dominated by the wild Tornionjoki stock,

while the catches on the Swedish side were more mixed, local stocks well presented in the catches.

In summary, mixed stock genetic analyses of Baltic Sea salmon catches performed over several years show that salmon from the AU 4–6 stocks do not occur in catches of Åland Islands and Gulf of Bothnian coastal salmon fisheries that take place only in summer. The WGBAST full life-history model (that currently includes only AU1–AU4 stocks) is constructed accordingly (see Stock Annex). In the Gulf of Finland commercial fishery, AU5 and AU6 fishes occur in small proportions in Estonian catches. In the main basin, AU5 and AU6 fish are also caught only occa-sionally in very small proportions by the commercial fisheries. The weak stocks of the Swedish rivers Ljungan and Emån have only been present in the catches near their respective river mouths. Fisheries in these weak rivers are presently closed.

4.5.3.2 A model predicting stock composition and catches of individual stocks in