Status of the assessment unit 1–4 stocks and development of fisheries in the

The full life-history model (FLHM) was run with two chains for 675 000 iterations after an adap-tive phase of 10 000 iterations. The first 150 000 iterations were discarded as burn-in and the chains were thinned with an interval of 350 to yield a final sample size of 3000 (1500 iterations from each of two chains). Inspection of traceplots and Gelman-Rubin diagnostics indicated poor convergence for many parameters. On closer inspection it became apparent that one of the chains (chain one) was getting stuck at implausible values for many variables. It was therefore decided to base results on only one chain for this year’s assessment as in 2019. In order to ensure that the most representative chain was selected for each parameter and variable in the model, the means from each chain were compared to posterior means from a longer converged run of 2020s assess-ment model, and the chain with the closest mean was selected for that parameter/variable. Start-ing with chain two as the default, this resulted in 9219 parameter/variables beStart-ing substituted in from chain one, out of a possible 19 606 in the longer converged run. Some caution must therefore be taken in the interpretation of results. In the text and figures that follow, medians and 90%

probability intervals (PI’s) are used where possible as statistics of posterior probability distribu-tions.

The results indicate a decreasing long-term trend in the post-smolt survival until mid-2000, after which survival has generally somewhat improved (Figure 4.2.3.1). The lowest overall survival was estimated for salmon that smolted in years 2005–2006 (median estimate around 8–10%

among wild and 5% among reared smolts), and survival was relatively low also in 2007–2009.

Low survivals were estimated for either wild or reared smolts also in some years of the last dec-ade, but the average survival in that decade was higher than in 2005–2009: 15% for wild smolts and 9% for reared smolts (median estimates ranging from 11–19% and 3–14% among wild and reared post-smolts, respectively). Survival was relatively high especially among wild salmon that smolted in 2010–2012 and 2014 (Figure 4.2.3.1). After the relatively high survival among 2017 wild smolts (16%) and poor survival among 2018 wild smolts (11%), survival is currently close to its average level (14% in 2019, which is the last smolting year with data to estimate).

The adult natural annual survival of wild salmon (median 91%, PI 86–95%) is estimated to be clearly higher than that of reared salmon (median 76%, PI 71–85%). Thus, the difference in total sea survival back to the spawning/stocking site for wild and reared salmon is large because of the survival difference both at post-smolt and at later marine stages.

Maturation (homing rate) of 1-sea winter salmon (grilse) has in most years been around 10–20%

(average of medians over the whole time-series is 16%) and 20–50% (average of medians over the whole time-series is 34%) among wild and reared individuals, respectively (Figure 4.2.3.2).

Differences between wild and reared salmon are smaller among multi-sea winter salmon, but in each sea age reared salmon has on average higher maturation rate. Generally, 30–60%, 60–70%

and 50–70% of 2SW, 3SW and 4SW feeding salmon have matured, respectively. The estimated maturation rates of four-sea winter are on average lower than those of three-sea winter salmon.

This is against intuition but might be an artefact due to the inconsistency between current model assumptions (no repeat spawners, all fish mature at latest after five sea winters) and the biology of salmon (some repeat spawners exist and some salmon have a longer lifespan than five years

at sea). Maturation rates of reared salmon have generally increased over time, but no similar trend is visible among wild salmon. Maturation rates were generally on the lowest levels around 2010–2012.

The full life-history model allows estimation of the stock-specific stock–recruit relationships, which are presented as summary statistics (Tables 4.2.3.1 and 4.2.3.2) and graphically (Figures 4.2.1.1, 4.2.3.3 and 4.2.3.4). Table 4.2.3.2 and Figure 4.2.3.4 also show the estimates of the stock-specific reference points (R^lim and R^MSY), which are used to assess stock status. “Equilibrium smolt production” corresponds to the Potential Smolt Production Capacity PSPC, i.e. the average smolt production that can be reached in the long term without fishing. It is important to note, that these PSPC estimates are not directly comparable to the PSPC estimates presented in earlier years’

assessments (e.g. ICES, 2019), where estimated PSPCs from the final year of the assessment pe-riod were used. In this year’s assessment, PSPCs from simulation are used (as in 2020), assuming reversion to long-term average vital rates (covering the whole historical time-series) for most time-varying parameters into the future. Among stocks the point values of Rlim and RMSY range from 15–40% and 60–85% of their corresponding PSPC’s point values, respectively (Table 4.2.3.2).

Figure 4.2.3.3 gives an indication of river-specific stock–recruit curves. The blue clouds in the figure panels indicate posterior probability distributions of all the historical estimates of yearly egg deposition and corresponding smolt abundance (the density of the cloud indicates the prob-ability). Curves added in the figure panels are draws from the posterior distribution of the Beverton–Holt stock–recruit function. Figure 4.2.3.4 illustrates how uncertainty related to the estimates of PSPC, R^lim and R^MSY vary between stocks. It is difficult to fully explain the between-stock variation in the level of uncertainty, but it is likely an outcome of several factors like between- stock-specific assumptions about vital rates, the amount of stock-stock-specific data, the coherence of data and the amount of contrast existing in the data in relation to the stock size. The total combined PSPC estimate containing all the AU 1-4 stocks is about 3.1 million (median, 90% PI’s 2.5-4.1 million) smolts (Table 4.2.3.2). Of this, AU 1 stocks account for about 80%, and AU 2 stocks ac-count for about 18%. When adding the point estimates of PSPC shown in the Table 4.2.3.3 for the AU 5 (301 000 smolts) and AU 6 (273 000 smolts), which are based of expert judgments, the total combined PSPC of all the assessed Baltic Sea salmon stocks is about 3.7 million smolts.

Since the mid-1990s, the status of many wild salmon populations in the Baltic Sea has improved, and the total wild production has increased from less than 0.5 to about three million smolts (Fig-ure 4.2.3.5, Table 4.2.3.3). After the record year 2017 (with median estimate of 3.14 million smolts) the total wild production has somewhat declined and it was 2.75 million smolts (median esti-mate) in 2020. Since the mid-2010s, the total smolt production of the AU 1 stocks has been clearly above the median estimates of both the combined Rlim and RMSY of AU 1, and it has been fluctu-ating close to the median estimate of combined PSPC (R⁰) of these stocks. In AU 2, the combined smolt production has been fluctuating around the median estimate of the combined R^MSY of AU 2 stocks. Also, in the AU 3 and AU 4 total smolt production has been recently near the median estimates of the combined R^MSY of the respective AU’s. Since the mid-2010s, the total combined AU 1–4 smolt production has been fluctuating between the median estimates of the total com-bined R^lim and R^MSY of all these AUs (Figure 4.2.3.5).

There are regional differences in trends in smolt production. For the wild salmon stocks of AUs 1–2, the very fast recovery of smolt production indicates high steepness for stock–recruit rela-tionships in these rivers. The recovery is most pronounced in the largest rivers, but recently the salmon stocks spawning in smaller ‘forest rivers’ of the region (Åbyälven, Rickleån, Sävarån, Öreälven, Lögdeälven) have speeded up their recovery. However, their stock status (current pro-duction level against MSY) is assessed to be lower than that of the larger salmon rivers, as dis-cussed below. The two wild stocks in AU 3 have also recovered, but the estimates of the current and/or the potential smolt production of Ljungan and Testeboån are highly uncertain. In AU 4

the Mörrumsån stock has stayed relatively stable, while the abundance in Emån has been grad-ually increasing. The AU 5 stocks are characterized by large interannual variation in smolt pro-duction and varying trends in the propro-duction. Smolt propro-duction in the Nemunas river system has been increasing especially in the latest years, while in Salaca and Gauja the production has been fluctuation without clear trends. Smolt production in Venta shows a decreasing trend.

Many AU 5 rivers are very small and their estimated PSPC is in some thousands of smolts only;

the existing data from these rivers are fragmentary and typically indicate zero or near-zero an-nual smolt production (see more details in Section 4.2.4).

By comparing the final year (2020) posterior smolt production (Table 4.2.3.3) against the esti-mated reference points Rlim and RMSY, it is possible to evaluate the current status of the AU 1–4 stocks in terms of their probability to reach the reference points (Table 4.2.3.4a). Table 4.2.3.4b contains wild and mixed AU 5–6 stocks, which are currently not included in the FLHM. These stocks have not been analytically derived, but expert judgments are used to classify their current status in relation to their PSPC. Because the estimates of annual smolt production vary greatly among AU 5–6 stocks (partly an artefact caused by assuming that all smolts are 2-year olds), the current status assessment is calculated in two ways: 1) by using only the 2020 smolt production estimate, and 2) by using the average of the 2018–2020 smolt production estimates.

Out of the 17 assessed stocks in AU 1–4, nine have reached Rlim with >95% probability, three stocks have reached R^lim with 70–95% probability, two stocks have reached R^lim with 50–70%

probability, and three stocks have reached R^lim with <50% probability (Table 4.2.3.4a). All stocks in the AU 1 are estimated to have reached their R^lim with 99–100% probability, and the corre-sponding probabilities of having reached their RMSY vary between 60–80%. In AU 2, three stocks (Piteälven, Byskeälven and Vindelälven) have reached their R^lim with >95% probability, while among the rest of the AU 2 rivers the corresponding probabilities range from 40% (Lögdeälven) to 89% (Sävarån). The probabilities for having reached R^MSY vary between 9% (Rickleån and Lö-gdeälven) and 77% (Piteälven). In AU 3, Ljungan has a low probability (< 40%) of having reached either of the reference points, while Testeboån has likely (with >70% probability) reached both reference points. A similar divided status prevails among the AU 4 stocks, where Mörrumsån has likely reached both R^lim (100% probability) and R^MSY (76% probability), whereas Emån has unlikely reached either of them (28% and 9% corresponding probabilities). As discussed in Sec-tion 4.4.2, current stock status of Piteälven (AU 2) and Testeboån (AU 3) is most likely overesti-mated.

Among the twelve AU 5 stocks, the wild Salaca (2020 and 2018–2010 average smolt production is 43% and 40% of PSPC, respectively) and mixed Nemunas (2020 and 2018–2010 average smolt production is 32% and 21% of PSPC, respectively) stocks have the highest current status. Among the remaining wild and mixed AU 5 stocks, current smolt production is <10% of the respective PSPC (Table 4.2.3.4b).

A large majority of the twelve AU 6 stocks has reached a higher proportion of their PSPC than the AU 5 stocks. Smolt production in the Kunda stock has reached 100% of its PSPC, and the production in the Keila stock is also very high (2020 and 2018–2010 average smolt production is 88% and 96% of PSPC, respectively). The current smolt production is <10% of the PSPC in the mixed stocks of Luga, Valgejögi, Jägala and Vääna (Table 4.2.3.4b).

The full life-history model (FLHM) captures quite well the overall historic fluctuation of catches in various fisheries, especially from the last ten years (Figure 4.2.3.6). However, catches from the first decade of this millennium tend to become underestimated for most of the years and fisher-ies. The model also does not fully capture the high river catches of the years 2008–2009 (Figure 4.2.3.6).

The model is fitted to the proportion of wild and reared salmon (separately for ages 2SW and 3SW) in the offshore catches. The posterior estimates of wild vs. reared proportions follow rather closely the observed proportions (Figure 4.2.3.7).

An increasing long-term trend in the number of spawners is seen in most of the rivers of the AUs 1–4 (Figure 4.2.3.18). Spawner abundance increased particularly in the years 2012–2014. In Simo-joki, the very high estimates of spawners around the turn of the millennium are a result of very intensive stocking of hatchery-reared parr and smolts in the river during the late 1990s. The model captures trends seen in fish ladder counts, even short-term variation in rivers where the data are not used for model fitting (e.g. Byskeälven). Annual variation in river conditions affect the success of fish to pass through ladders and, therefore, the ladder counts themselves are not ideal indices of spawner abundance.

In Kalixälven, Åbyälven and Rickleån the development of spawner abundance estimated by the model appears more optimistic than the development observed in the fish ladder counts. In Ka-lixälven, the counter is located about 100 km from the river mouth with large spawning areas downstream. In Åbyälven and Rickleån fish ladders were constructed around the turn of the millennium and salmon are gradually repopulating the upstream sections above the dams.

Therefore, counts in these rivers account for a small fraction of the total spawner population, and the counts may not well represent the actual development of these river stocks.

Unlike in the other AU 1–3 stocks, the amount of spawners dramatically dropped in Ume/Vindelälven for the years 2015–2018. Since 2014, the fish ladder counts in this river have not been as low as the model estimated numbers of spawners (Figure 4.2.3.8 vs. Table 3.1.1.2 and Figure 3.1.1.3). This is due to the need to accommodate Ume/Vindel stock dynamics in the FLHM to the extra losses among female salmon to reach spawning grounds in this river (see Section 4.2.1 and Stock Annex, Section C.1.9). The drop in spawner abundance in Ume/Vindelälven is dramatically decreasing the current and near-future smolt production (Table 4.2.3.3 and Figure 4.2.3.8b). However, the most recent (2019–2020) spawning runs into the river have been abun-dant and the smolt production is expected to increase rapidly starting 2022.

The general synchronous drops and increases in the observed spawner counts are well-captured by the model, also the most recent drop observed from 2016 to 2017–2018. This is probably a consequence of fitting the model to spawner counts in combination with assuming annually var-ying maturation rates; maturation rates are estimated to be lower preceding poor spawning runs and higher preceding high spawning runs (Figure 4.2.3.2 vs. Figure 4.2.3.8). Also, the effect of annually varying post-smolt survival is visible in spawner counts and estimates, e.g. the low survival of the 2016 smolt cohort contribute to the low spawner abundance especially in 2018.

For 2021, the FLHM predicts moderate spawner abundance in most rivers. This prediction must, however, be taken with caution, because the prediction is very uncertain, and e.g. natural condi-tions at sea during the spring 2021 (not currently well known/predicted) are expected to modify the spawning run strength via maturation rates and run timing.

Despite some fluctuations, there was a strong long-term decreasing trend in the harvest rate of driftnets until the total ban of this gear type in 2008 (Figure 4.2.3.9a). The harvest rate of longlin-ing has been fluctuatlonglin-ing a lot (between less than 0.05 to about 0.3 among MSW salmon). After the peaks in 2003–2005 and again in 2011, this harvest rate dropped to about 0.1 and for the two last years (2019–2020) the harvest rate dropped further to below 0.05. The harvest rate in trolling increased from the 1990s until 2007–2010, when it was 0.07–0.08 (Figure 4.2.3.9b). During the last decade this harvest rate has been on the level of 0.03–0.05. The combined offshore harvest rate (driftnetting, longlining and trolling) shows a clearly decreasing trend from about 0.5 in the early 1990s to below 0.1 in the last two years (Figure 4.2.3.10). Since the early 2000s, the coastal harvest rate, which predominantly consists of trapnet fishing, has decreased almost continuously (Figure 4.2.3.9c). Currently the harvest rate of this fishery is about 0.15 for the AU 1 salmon (which has

the highest coastal harvest rate of all Baltic salmon) (Figures 4.2.3.9c and 4.2.3.10). Estimates of harvest rates in the rivers are inaccurate and lack a clear trend (Figure 4.2.3.9d). River-specific data indicate that there can be substantial variation in the harvest rate between rivers (Section 3.1), which is currently not taken into account in the FLHM.