Access the full text.
Sign up today, get DeepDyve free for 14 days.
INTRODUCTIONModern commercial fishing targets particular fish species and size, avoiding unwanted species and harm to the environment. Although advanced fishing techniques increasingly enable selective fishing, unwanted bycatch still commonly occurs (Uhlmann & Broadhurst, 2015). For a fishery to be environmentally sustainable, the survival rate of the bycatch released must be high, and tight legislation is frequently established to accomplish the goals (e.g., EU, 2021; STECF, 2020).In addition to commercial fisheries, recreational fishing may also include releasing the catch for a variety of reasons (Arlinghaus et al., 2007). Especially some iconic species, like Atlantic salmon (Salmo salar), have increasingly been studied to understand the effects of catch‐and‐release practices in different types of fisheries. Sources of potential problems and recommended methods to be used are being intensively studied (e.g., Lennox et al., 2017).The commercial fisheries for Atlantic salmon in the Baltic Sea occur mainly along the coasts of Finland and Sweden using trap nets (International Council for the Exploration of the Sea [ICES], 2021). Total allowable catch of Baltic salmon is annually stated in EU Council Regulation based on scientific advice (ICES, 2021). A landing obligation has been set for the quota‐managed species in EU fisheries, but there have been exemptions for Baltic salmon fisheries, based on the proven high survival after release from trap nets (EU, 2021; STECF, 2020). In Finland, commercial salmon fishing is strictly regulated by strict spatial and temporal measures, and only trap nets are allowed in coastal fishing. The national legislation currently in force does not allow selection of salmon in targeted fishery, and trap nets should be lifted when quota per fisher is reached. However, releasing of fish may be applied if quota is reached during the trap emptying, or salmon are a bycatch when targeting other species.In recent years, there have been concerns about potentially higher release mortalities of salmon (Östergren et al., 2020) than shown in studies conducted in the northern Baltic Sea in the early 2000s (Rivinoja et al., 2001; Siira et al., 2006). The suggested reasons for the higher mortalities documented include the different trap net designs used (Hemmingsson et al., 2008), different trap net hauling techniques (Östergren et al., 2020), the general health problems of Baltic salmon (Weichert et al., 2021), or their combinations with varying environmental factors (e.g., a high water temperature in rivers and estuaries).Fishing gear and operations can be developed to minimise unwanted mortality (Suuronen et al., 2012; Uhlmann & Broadhurst, 2015). Swedish studies on Baltic salmon have suggested that an air‐operated pontoon trap with an emptying chute may cause more fish injuries and higher releasing mortality than a pontoon trap equipped with a lifting bag, which was developed to decrease injuries for fish during trap net emptying (Östergren et al., 2020). Ruokonen et al. (2020) found some evidence that salmon caught with the pontoon trap with a lifting bag had less external damage than salmon caught with a trap equipped with an emptying chute.The estimation of release mortality is a complex task because several sources of mortality and affecting factors can be involved in the process, depending on the methods used (Breen & Catchpole, 2021; Patterson et al., 2017). Tagging‐recapture methods are often used to assess discard or release mortality, but data analysis often lacks information about sources of uncertainty related to the process. The best option would be to integrate tagging with captive observations or vitality assessments (Breen & Catchpole, 2021). However, the use of parallel approaches is not always possible. For such cases, Bayesian inference offers a possibility to combine existing information from various sources. The results of Bayesian models are presented as probability distributions, and they are therefore relatively easy to interpret even if a complex model framework is established (e.g., Michielsens et al., 2008). The utilisation of prior knowledge is beneficial, for example, when there is a need to estimate the relative magnitude of different sources of mortality related to handling and releasing fish and other factors related to the selected assessment method.We designed the first rigorous tagging experiment that compared the effect of two emptying techniques of a pontoon trap on the mortality of released Baltic salmon. Compared to previous studies (Östergren et al., 2020; Ruokonen et al., 2020), the experimental design in our study was specifically designed to elucidate the effect of the emptying technique. The study was conducted by tagging salmon in the summer of 2021 in the Bothnian Sea during their spawning migration. Fish were transferred to the boat using the two emptying techniques currently used in the Baltic coastal salmon fisheries. The external injuries of salmon were recorded visually during tagging, and the types and probability of getting an injury were compared between the two emptying methods.A custom‐built Bayesian system model was employed to estimate release mortalities caused by the two emptying methods. Another Bayesian model was used to estimate emptying method specific probabilities for the fish to get specific types of injuries. In the release mortality model, a wide range of prior information was used containing fishing and natural mortalities from the stock assessment of Baltic salmon, tag loss rates, and handling mortality from previous studies as well as expert elicitation on tag reporting rates.MATERIALS AND METHODSStudy area, fish tagging, and data collectionThe study was conducted in collaboration with four commercial fishers on the coast of the Bothnian Sea in Merikarvia, Finland (ICES subdivision 30), between late May and early July 2021. In the traditional emptying technique, salmon are transferred from the pontoon trap to a water‐filled tank on board a boat using a fibreglass emptying chute. In this technique, the salmon are out of the water during trap lifting. In the other method, salmon are lifted into a water tank on board with the aid of a lifting bag made of knotless mesh netting. As the fish chamber of a pontoon trap is raised towards the surface during the lifting, fish slide into the submersed lifting bag, which can then be lifted onto the boat.All the fishers had two pontoon traps with a similar basic design (Hemmingsson et al., 2008), and the traps were in the same area to ensure comparability between emptying methods. One of the traps of each fisher had a traditional emptying chute made of fiberglass, and the other trap was fitted with a lifting bag made from knotless netting (6 m long, mesh size 25 mm) (Figure 1). Traps with a lifting bag were emptied by lifting the bag with the salmon catch onto the boat by hand. If the catch was large, fish were lifted in two or three batches.1FIGUREEmptying a pontoon trap using a fibreglass emptying chute (left) and a lifting bag (right).The salmon caught were emptied into a tarpaulin tank filled with water and tagged by trained personnel of Natural Resources institute Finland (Luke) with an arrow tag (type PDA, bright green, length 15 cm; Hallprint Ltd.) at the base of the dorsal fin. Prior to tagging, the total length of each fish was measured. Fish tagging was conducted following national guidelines on animal welfare, and no specific ethic statement was needed for the study. A reward of €25 was offered for each tag recovery to all fishers in coastal areas and rivers in Finland and Sweden. Only wild salmon with an intact adipose fin were tagged because all stocked salmon in Finland and Sweden have been fin‐clipped since 2017. The seawater temperature rose from 8°C to 17°C during the tagging period (Figure 2).2FIGUREThe number of tagged salmon on the coast of the Bothnian Sea, Finland, in the summer of 2021, caught with different pontoon trap types (emptying chute or lifting bag) and the weekly mean water temperature recorded during tagging.In total, 202 salmon caught with a pontoon trap with an emptying chute and 211 salmon from PU traps with a lifting bag were tagged (Figure 2). All the tagged fish, except one, were multi‐sea‐winter (MSW) salmon. The size distributions of the tagged fish were similar in both tagging groups (Figure 3).3FIGURESize distributions of salmon tagged from traps with an emptying chute (a) and a lifting bag (b). The dashed line indicates the mean lengths of tagged salmon in both groups.The condition of each tagged salmon was visually inspected, and detected injuries were recorded. Injuries were categorised as follows: scale loss = loss of scales divided into three categories (no scale loss [<5%], minor scale loss [5%–15%], major scale loss [>15%]); fin injury = split fin or other injury; red eye = internal bleeding in the eye; mesh injury = wounds and scratches to the skin caused by meshing of fish through the netting; wounds = wounds, bites, and scratches by seals or other unidentified factors; and mouth/jaw injury = mouth/jaw injury (e.g., ripped maxilla) caused by fishing gear.Statistical modellingThe release mortality of salmon from pontoon traps with two different emptying methods was evaluated using a custom‐built Bayesian system model. The method was chosen because it allows a combination of different data sources, and the results are intuitively interpretable as probabilities. In Bayesian models, prior information is defined as prior distributions, which are updated through other model assumptions and observed data sets to posterior distributions. To assess the magnitude of different sources of mortality based on tag returns, the use of prior information on fishing and natural mortality is necessary. The system model describes the partially unobservable processes of migration, mortality, survival, and capture of tagged and released fish.Probability modelThe probability model comprises a time span of 15 weeks, which corresponds to the period from the first tagging batch to the last date on which tagged fish have been reported to be caught. Tag recaptures are divided into two categories: recaptures (1) from coastal fisheries and (2) from river fisheries. River fisheries refer to all salmon rivers on the Finnish and Swedish sides of the Bothnian Bay to which tagged salmon may have been migrating.In the model structure, we assume that the tagging and release mortalities take place immediately after releasing. This assumption is made for computational simplicity, although in practice these mortalities are also highly likely to occur later during the migration. We therefore assume that the number of tagged salmon alive at sea after tagging mortality and release mortality induced by trap type j (Nt,jalive$N_{t,j}^{alive}$) is binomially distributed:1Nt,jalive=Bin(pjalive,Nt,jtagged),$$\begin{equation}N_{t,j}^{{\rm{alive}}} = {\rm{Bin}}(p_j^{{\rm{alive}}},\ N_{t,j}^{{\rm{tagged}}}\ ),\end{equation}$$where Nt,jtagged$N_{t,j}^{{\rm{tagged}}}$ is the number of tagged salmon released from a trap type j during week t. Index j indicates the type of trap: j=1$j\ = \ 1$ corresponding to pontoon trap with an emptying chute and j=2$j\ = \ 2$ corresponding to a pontoon trap with a lifting bag. Furthermore, pjalive$p_j^{{\rm{alive}}}$ is the probability that a salmon released from a trap type j survives from the tagging and releasing mortalities:2pjalive=exp−(hj+ipmark).$$\begin{equation}p_j^{{\rm{alive}}} = \ {\rm{exp}}\left( { - ({h}_j + i{p}^{{\rm{mark}}})} \right).\end{equation}$$Here hj${h}_j$ is the instantaneous trap induced mortality of trap type j, and ipmark$i{p}^{{\rm{mark}}}$ is the instantaneous mortality caused by the process of tagging and handling, regardless of the trap type.Thus, the probability of dying because of release mortality induced by trap type j is derived as:3pjdie=1−exp−(hj+ipmark·qhand),$$\begin{equation}p_j^{{\rm{die}}} = \ 1 - {\rm{exp}}\left( { - ({h}_j + i{p}^{{\rm{mark}}} \cdot {q}^{{\rm{hand}}})} \right),\end{equation}$$where qhand expresses the share of mortality caused by handling of total mortality caused by the process of tagging and handling, that is, excluding the tagging mortality. This division must be made because all the released salmon in the experiment are tagged.Furthermore, we denote the initial number of tagged salmon at sea available for coastal fisheries with:4n1,1,j=N1,jalive.$$\begin{equation}{n}_{1,1,j} = N_{1,j}^{{\rm{alive}}}.\end{equation}$$The first index of variable n indicates the area, later denoted with i=1${\bf{i\ }} = \ 1$ for sea and i=2${\bf{i\ }} = \ 2$ for river. The initial number of tagged salmon in river areas is set as zero,5n1,2,j=0,$$\begin{equation}{{\bf{n}}}_{1,2,{\bf{j}}} = \ 0,\end{equation}$$indicating that no tagging takes place at rivers.We assume that during a single week of the study period, a tagged salmon could either die, survive, and stay at sea or survive and move to a river. The fractions that survive are denoted as:6q1SS=p1surv·pkeep·1−pmoveq2SS=p2surv·pkeepqSM=p1surv·pkeep·pmove,$$\begin{equation}\left\{ { \def\eqcellsep{&}\begin{array}{@{}*{1}{c}@{}} {\ {\bf{q}}_1^{{\rm{SS}}} = \ {\bf{p}}_1^{{\rm{surv}}} \cdot {{\bf{p}}}^{{\rm{keep}}} \cdot \left( {1 - {{\bf{p}}}^{{\rm{move}}}} \right)}\\ {\ {\bf{q}}_2^{{\rm{SS}}} = \ {\bf{p}}_2^{{\rm{surv}}} \cdot {{\bf{p}}}^{{\rm{keep}}}}\\ {\ {{\bf{q}}}^{{\rm{SM}}} = \ {\bf{p}}_1^{{\rm{surv}}} \cdot {{\bf{p}}}^{{\rm{keep}}} \cdot {{\bf{p}}}^{{\rm{move}}}} \end{array} } \right.,\end{equation}$$where q1SS${\bf{q}}_1^{{\rm{SS}}}$ is the fraction that survives and stays at the sea, qSM is the fraction that survives and moves from the sea to a river, and q2SS${\bf{q}}_2^{{\rm{SS}}}$ is the fraction that survives and stays in the river. The survival probability for 1 week in area i is denoted with pisurv${\bf{p}}_{\bf{i}}^{{\rm{surv}}}$, whereas the parameter pmove denotes the probability of moving from the sea to a river and pkeep the probability that a tag will not be lost during a single week.Thus, the total number of tagged salmon alive at sea during week t+1${\bf{t}} + 1$ originally released from a trap of type j can be calculated as:7nt+1,1,j=nt,1,j·q1SS+Nt+1,jalive.$$\begin{equation}{{\bf{n}}}_{{\bf{t}} + 1,\ 1,\ {\bf{j}}} = {{\bf{n}}}_{{\bf{t}},1,{\bf{j}}} \cdot {\bf{q}}_1^{{\rm{SS}}} + {\bf{N}}_{{\bf{t}} + 1,{\bf{j}}}^{{\rm{alive}}}.\end{equation}$$Similarly, the total number of tagged salmon alive in rivers during week t+1${\bf{t}} + 1$ originally released from a trap of type j is calculated as:8nt+1,2,j=nt,1,j·qSM+nt,2,j·q2SS.$$\begin{equation}{{\bf{n}}}_{{\bf{t}} + 1,\ 2,\ {\bf{j}}} = {{\bf{n}}}_{{\bf{t}},1,{\bf{j}}} \cdot {{\bf{q}}}^{{\rm{SM}}} + {{\bf{n}}}_{{\bf{t}},2,{\bf{j}}}\ \cdot {\bf{q}}_2^{{\rm{SS}}}.\end{equation}$$The number of recaptured salmon from area i during week t that were originally released from a trap of type j(Nt,i,jrecap)$({\bf{N}}_{{\bf{t}},{\bf{i}},{\bf{j}}}^{{\rm{recap}}})$ is assumed to be Poisson distributed:9Nt,i,jrecap∼PoissonHRi,j·nt,i,j·rireport+0.001,$$\begin{equation}{\bf{N}}_{{\bf{t}},{\bf{i}},{\bf{j}}}^{{\rm{recap}}}\sim \ {\rm{Poisson}}\left( {{\rm{HR}}_{{\bf{i}},{\bf{j}}} \cdot {{\bf{n}}}_{{\bf{t}},\ {\bf{i}},{\bf{j}}} \cdot \ {\bf{r}}_{\bf{i}}^{{\rm{report}}} + 0.001} \right),\end{equation}$$where HRi,j${\rm{HR}}_{{\bf{i}},{\bf{j}}}$ is the harvest rate of salmon originally released from a trap of type j at area i and rireport${\bf{r}}_{\bf{i}}^{{\rm{report}}}$ is the reporting rate of tags at area i. A small constant is added for numerical stability. In the following step, informative prior distributions for natural and fishing mortalities in different areas, as well as for the probability that a tag is not lost, are defined and transformed into instantaneous mortalities for a 1‐week period. With these parameters, we can calculate the harvest rates of tagged salmon:10HRi,j=pkeep·1−pi,jsurv·FiMi+Fi,$$\begin{equation}{\rm{HR}}_{{\bf{i}},{\bf{j}}} = {{\bf{p}}}^{{\rm{keep}}}\ \cdot \left( {1 - {\bf{p}}_{{\bf{i}},{\bf{j}}}^{{\rm{surv}}}} \right) \cdot \frac{{{{\bf{F}}}_{\bf{i}}}}{{{{\bf{M}}}_{\bf{i}} + {{\bf{F}}}_{\bf{i}}}},\end{equation}$$where Fi${{\bf{F}}}_{\bf{i}}$ is the instantaneous fishing mortality and Mi${{\bf{M}}}_{\bf{i}}$ the instantaneous natural mortality at area i. Furthermore, the area‐specific probability of surviving from natural and fishing mortalities is calculated as:11pisurv=exp−Mi+Fi.$$\begin{equation}{\bf{p}}_{\bf{i}}^{{\rm{surv}}} = \ {{\bf exp}}\left( { - \left( {{{\bf{M}}}_{\bf{i}} + {{\bf{F}}}_{\bf{i}}} \right)} \right).\end{equation}$$Informative priors for causes of death and other types of fateAs described above, the tagging and trap‐induced release mortalities are assumed to take place immediately after releasing. After this, a tagged salmon can be exposed to any of the following events during a single week in the model: death because of natural or fishing mortality, the loss of a tag, and movement from the sea to a river. Salmon that migrate to rivers are assumed to remain in the river and be exposed to the previously mentioned events each week until the end of the study period. In addition, tags from recaptured salmon can either be reported or remain unreported.We derived informative priors for fates of tagged salmon from several literature sources. The prior distributions for natural and fishing mortalities were derived by averaging posterior estimates of the Baltic salmon stock assessment model outcome between 2018 and 2020 (ICES, 2021). For example, the prior distribution for instantaneous mortality at sea during 11 weeks of coastal fishery was approximated as:12Fsea∼logN−1.86,0.1312,$$\begin{equation}{{\bf{F}}}^{{\rm{sea}}}\sim {\rm{log}}{\bf{N}}\left( { - 1.86,{{0.131}}^2} \right),\end{equation}$$which was further transformed into instantaneous mortality for 1 week as:13F1=Fsea/11.$$\begin{equation}\ {{\bf{F}}}_1 = \ {{\bf{F}}}^{{\rm{sea}}}/11.\end{equation}$$Similarly, we derived the prior distributions for instantaneous mortality at river fishery (Friver), natural mortality at sea (Msea), and natural mortality in the river (Mriver) as:14Friver∼logN−1.60,0.3242Msea∼logN−1.87,0.3132Mriver∼logN−2.38,0.3122,$$\begin{equation}\left\{ { \def\eqcellsep{&}\begin{array}{@{}*{1}{c}@{}} {{{\bf{F}}}^{{\rm{river}}}\sim \ {\rm{log}}N\left( { - 1.60,\ {{0.324}}^2} \right)\ }\\ {{{\bf{M}}}^{{\rm{sea}}}\sim \ {\rm{log}}N\left( { - 1.87,{{0.313}}^2} \right)}\\ {{{\bf{M}}}^{{\rm{river}}}\sim \ {\rm{log}}N\left( { - 2.38,{{0.312}}^2} \right),} \end{array} } \right.\end{equation}$$which were again transformed into instantaneous mortalities for 1 week:15F2=Friver/12M1=Msea/8M2=Mriver/52.$$\begin{equation}\left\{ { \def\eqcellsep{&}\begin{array}{@{}*{1}{c}@{}} {\ {{\bf{F}}}_2 = \ {{\bf{F}}}^{{\rm{river}}}/12\ }\\ {{{\bf{M}}}_1 = \ {{\bf{M}}}^{{\rm{sea}}}/8}\\ {\ {{\bf{M}}}_2 = {{\bf{M}}}^{{\rm{river}}}\ /52.} \end{array} } \right.\end{equation}$$The number of weeks that each type of mortality is assumed to take place reflects the period of impact in the Baltic salmon stock assessment model (ICES, 2021). In the model, the length of the fishing season is assumed to be 11 weeks for sea fisheries and 12 weeks for river fisheries. Both natural mortalities at rivers and at sea are, in fact, based on the same estimate of annual mortality that is equal to Mriver${{\bf{M}}}^{{\rm{river}}}\ $ (ICES, 2021). Thus, the natural mortality in rivers during 1 week can be calculated by dividing the annual estimate by 52. The Msea is essentially the same estimate as the Mriver, but it has been increased slightly because of increased mortality due to seal predation at coastal areas. Therefore, Msea is somewhat higher than the annual natural mortality without the seal predation (called Mriver here). The natural mortality at coastal sea areas during 1 week is obtained by dividing Msea by 8 weeks, which corresponds to the assumed period it takes for salmon to migrate from the Baltic main basin into the rivers.For tagging related events, informative prior distributions were derived from the study of Siira et al. (2006). Based on their analysis on tag retention rate, we assume that the probability that a tagged salmon loses the tag during a 3‐month period is Beta‐distributed:16plose∼Beta12,68.$$\begin{equation}{{\bf{p}}}^{{\rm{lose}}}\sim {\rm{Beta}}\left( {12,68} \right).\end{equation}$$We transformed this probability into instantaneous mortality for 1 week as:17ipkeep=−log1−plose/12$$\begin{equation}{\bf{i\ }}{{\bf{p}}}^{{\rm{keep}}} = \ - {{\bf log}}\left( {1 - {{\bf{p}}}^{{\rm{lose}}}/12} \right)\end{equation}$$and again, back into a proportional scale as:18pkeep=exp−ipkeep.$$\begin{equation}{{\bf{p}}}^{{\rm{keep}}} = \ {{\bf exp}}\left( { - \ {\bf{i}}{{\bf{p}}}^{{\rm{keep}}}} \right).\end{equation}$$Furthermore, we utilised the data from the tank experiment of Siira et al. (2006) to derive the prior distribution for the proportion of mortality caused by handling of total mortality caused by the process of tagging and handling, denoted by qhand. We use the same procedure as in Ruokonen et al. (2020) to derive qhand based on data from the tank experiment. As a result, we formulate the prior distribution as:19qhand∼logN−0.723,0.3712T,1.$$\begin{equation}{{\bf{q}}}^{{\rm{hand}}}\sim {\rm{log}}{\bf{N}}\left( { - 0.723,\ {{0.371}}^2} \right){\bf{T}}\left( {,1} \right).\end{equation}$$The T(, 1) notation indicates that the distribution is truncated to allow values up to a maximum of 1.The prior distribution for the probability of death because of tagging or releasing (pmark) during a 3‐month period is derived from the previous analysis by Ruokonen et al. (2020):20pmark∼Beta1.8,7.2.$$\begin{equation}{{\bf{p}}}^{{\rm{mark}}}\sim {\rm{Beta}}\left( {1.8,7.2} \right).\end{equation}$$This can be transformed into an instantaneous scale for 1 week as:21ipmark=−log1−pmark12.$$\begin{equation}{\bf{i}}{{\bf{p}}}^{{\rm{mark}}} = \ - \log \left( {1 - \frac{{{{\bf{p}}}^{{\rm{mark}}}}}{{12}}} \right).\end{equation}$$In addition to informative priors, we gave an uninformative prior distribution for the movement probability from sea to river:22pmove∼Unif0,1.$$\begin{equation}{{\bf{p}}}^{{\rm{move}}}\sim {\rm{Unif}}\left( {0,1} \right).\end{equation}$$Uninformative prior distribution was also assumed for trap‐induced mortality as23hj∼Unif0,12.$$\begin{equation}{{\bf{h}}}_{\bf{j}}\sim {\rm{Unif}}\left( {0,12} \right).\end{equation}$$All prior distributions are listed in Table 1.1TABLEList of symbols, descriptions, and prior distributions.SymbolDescriptionPrior distributionIndicestWeek from 1 to 15–iArea; 1: sea, 2: river–jTrap type; 1: pontoon trap with an emptying chute, 2: pontoon trap with a lifting bag–DataNt,jtagged${\bm{N}}_{{\bm{t}},{\bm{j}}}^{{\rm{tagged}}}$Number of tagged salmon released in week t from trap of type j–Nt,i,jrecap${\bm{N}}_{{\bm{t}},\ {\bm{i}},{\bm{j}}}^{{\rm{recap}}}$Number of tagged salmon released from trap type j recaptured in week t at area i–Latent variablesNt,jalive${\bm{N}}_{{\bm{t}},{\bm{j}}}^{{\rm{alive}}}$Number of tagged salmon released from a trap of type j alive in week t–pjalive${\bm{p}}_{\bm{j}}^{{\rm{alive}}}$Probability of a salmon released from a trap of type j surviving from the tagging and releasing mortalities–pjdie${\bm{p}}_{\bm{j}}^{{\rm{die}}}$Probability of a salmon dying because of release mortality induced by trap type j–nt, i, jNumber of tagged salmon released from trap type j that are available for fisheries at area i in week t–qiSS${\bm{q}}_{\bm{i}}^{{\rm{SS}}}\ $Surviving fraction of tagged salmon that stay at area i–qSMSurviving fraction of tagged salmon that move from sea to river–FiInstantaneous fishing mortality for 1 week at area i–MiInstantaneous natural mortality for 1 week at area i–HRi, jHarvest rate of tagged salmon originally released from a trap of type j at area i–pisurv${\bm{p}}_{\bm{i}}^{{\rm{surv}}}$Probability that a salmon released from a trap of type j survives from natural and fishing mortalities for 1 week at area i–rireport${\bm{r}}_{\bm{i}}^{{\rm{report}}}$Tag reporting rate at area i–pkeepProbability of keeping a tag for 1 week–ipkeepProbability of keeping a tag for 1 week at instantaneous scale–ipmarkInstantaneous mortality due to tagging and releasing–ParametershjTrap induced mortality for trap type jU(0,12)pmoveProbability of moving from sea to river for 1 weekU(0,1)ploseProbability of a tag loss for 12 weeksBeta(12,68)qhand${q}^{hand}$Proportion of mortality caused by handling of total mortality caused by tagging and handlinglogN(0.723, 0.3712)T(,1)pmarkProbability of dying because of tagging mortality or release mortality for 12 weeksBeta(1.8,7.2)FseaInstantaneous mortality at coastal fisheries for 11 weekslogN(−1.86,0.1312)FriverInstantaneous mortality at river fisheries for 12 weekslogN(−1.60,0.3242)MseaInstantaneous natural mortality on coast for 8 weekslogN(−1.87,0.3132)MriverInstantaneous natural mortality in rivers for 52 weekslogN(−2.38,0.3122)Ek,i${E}_{k,i}$Expert k ‘s prior for reporting rate at area iSee Table 2Expert‐elicited priors for reporting ratesThe reporting rates of recaptured tags are an essential piece of information when data from a mark‐recapture study are analysed. Unfortunately, good data on the level of reporting seldom exist, and eagerness to report may vary in time between areas and groups of fishers. We chose the approach of expert elicitation to formulate priors for reporting rates at coastal and river fisheries. Three experts best familiar with these fisheries were interviewed independently from each other. The experts formulated their views on the reporting rates in terms of either modal or median value, and in terms of an interval covering either 90% or 95% of the probability mass. The experts were also asked to include their views of the differences in reporting activity in Finnish and Swedish fisheries in their assessments.The expert views on the central tendency and the uncertainty of the reporting rates at sea and at river fisheries are shown in Supporting Information. A beta distribution was fitted for each distribution provided (Supporting Information; Table S1 and Figure S1). We used the method of Bayesian model averaging (BMA; Hoeting et al., 1999) to average the distributions provided by each expert to formulate prior distribution for reporting rates at area i:24rireport=EY,i,$$\begin{equation}{\bf{r}}_{\bf{i}}^{{\rm{report}}} = \ {{\bf{E}}}_{{\bf{Y}},{\bf{i}}},\end{equation}$$where Y follows a categorical distribution:25Y∼catw1:3.$$\begin{equation}{\bf{Y}}\ \sim {\rm{cat}}\left( {{\bf{w}}\left[ {1:3} \right]} \right).\end{equation}$$Equal weights w[1:3]=1/3${\bf{w\ }}[ {1:3} ] = \ 1/3$ were assigned for each expert.Injury modelTo estimate the probability that a salmon captured by either a pontoon trap with an emptying chute or one with a lifting bag will get an injury of a specific type, we fitted a small Bayesian beta‐binomial model.We assume that the observed number of salmon that have an injury i after being captured by a trap of type j (xi,j)${{\bf{x}}}_{{\bf{i}},{\bf{j}}})$ follows a binomial distribution:26xi,j∼Binpi,j,Nj,$$\begin{equation}{{\bf{x}}}_{{\bf{i}},{\bf{j}}}\sim {\rm{Bin}}\left( {{{\bf{p}}}_{{\bf{i}},{\bf{j}}},\ {{\bf{N}}}_{\bf{j}}} \right),\end{equation}$$where Nj${{\bf{N}}}_{\bf{j}}$ is the total number of salmon captured with a trap type j. Furthermore, we assume that the probability of an injury i${\bf{i}}\ $when captured by a trap of type j (pi,j${{\bf{p}}}_{{\bf{i}},{\bf{j}}}$) follows an uninformative beta‐distributed prior:27pi,j∼Beta1,1.$$\begin{equation}{{\bf{p}}}_{{\bf{i}},{\bf{j}}}\sim {\rm{Beta}}\left( {1,1} \right).\end{equation}$$Details of the model runsPosterior distributions for the release mortality model were estimated with the JAGS software by running a sample of 200,000 iterations with two chains (burn‐in 1000 iterations). Posterior distributions for the injury model were estimated correspondingly with a sample of 10,000 iterations with two chains (burn‐in 1000 iterations). The models’ convergence was ensured both visually (trace plots) and numerically (Rubin–Gelman diagnostics). The full model code for the release mortality model can be found in Appendix A and for the injury model in Appendix B.RESULTSTag observationsBy the end of December 2021, a total of 59 (14%) tags were returned, with a higher proportion from salmon released from pontoon traps with a lifting bag (Table 2, Figure 4). A higher proportion (59%) of tag observations was received from rivers than from coastal fisheries (Table 2, Figure 4).2TABLETag recoveries by trap type and location.Pontoon trap typeTaggedSeaRiverTotalPercentageChute20210142412Lifting bag21114213517Total413243559144FIGURETagging and tag recovery locations. International Council for the Exploration of the Sea [ICES] sub‐divisions are marked by numbers.Bayesian estimates for release mortalityEstimated posterior results (means and 90% highest probability density intervals, HDI) for parameters, as well as the corresponding statistics from prior distributions, are expressed in numerical form in Table 3 and as density graphs in Figure 5. The posterior distributions of fisheries mortality (Fsea and Friver) changed markedly from the given priors, whereas the posteriors of natural mortality (Msea and Mriver) were very close to the priors. This is understandable because the tag recapture data are unlikely to carry information about the natural mortality at sea or in rivers. The posterior distributions of parameters for which uninformative prior distributions were assigned (h1, h2, pmove) were updated from the corresponding priors.3TABLEPrior and posterior means and 90% highest density probability intervals (HDI) for parameters of release mortality model.ParameterPrior meanPrior 90% HDIPosterior meanPosterior 90% HDIqhand0.5[0.23,0.78]0.5[0.23,0.78]r1report$r_1^{{\rm{report}}}$0.73[0.57,0.9]0.82[0.69,0.96]r2report$r_2^{{\rm{report}}}$0.69[0.47,0.89]0.72[0.54,0.88]h15.96[0.00,20.79]0.25[0.00,0.5]h26.00[0.67,11.44]0.1[0.00,0.22]Fsea0.16[0.12,0.19]0.18[0.15,0.11]Friver0.21[0.1,0.32]0.27[0.17,0.38]Msea0.16[0.08,0.24]0.16[0.08,0.23]Mriver0.10[0.05,0.14]0.10[0.05,0.15]pmove0.5[0.01,0.91]0.29[0.19,0.39]pmark0.2[0.01,0.39]0.09[0.00,0.17]plose0.15[0.09,0.21]0.15[0.08,0.21]5FIGUREPrior (narrow line) and posterior (thick line) distribution for parameters of the Bayesian release mortality model.Parameters with informative prior distributions from the tank experiment of Siira et al. (2006), plose and qhand, had posterior estimates that closely resembled their priors. This makes sense because these informative distributions were derived from the tank experiment, as we did not gather data about the tag retention rate or tagging mortality in our study. The posterior distribution for the probability of dying because of tagging or releasing (pmark) was updated towards lower values from the prior, which was set up based on our previous study (Ruokonen et al., 2020). The posterior estimate of the tag reporting rate at sea fisheries (r1report${\bf{r}}_1^{{\rm{report}}}$) was updated into somewhat higher values from the prior, whereas the posterior in the tag reporting rate at river fisheries r2report${\bf{r}}_2^{{\rm{report}}}$ was quite similar to its prior.The posterior estimate for the probability of salmon dying due to release from a pontoon trap with an emptying chute was 24% on average (with 90% HDI 3%–43%) (Figure 6). The estimate for salmon released from a trap with a lifting bag was 13% on average (90% HDI 1%–24%) (Figure 6). Based on these posterior distributions, it can be stated with 78% probability that the mortality rate of salmon released from a trap with an emptying chute was higher than from a trap with a lifting bag. The release mortality from a trap with an emptying chute was on average 2.8 times higher than the release mortality from a trap with a lifting bag.6FIGUREThe estimated posterior distributions for probability of a salmon dying due to release from a pontoon trap with an emptying chute or with a lifting bag.InjuriesBased on a visual external examination, the tagged salmon were in a good general condition. No acute immediate mortality was observed during the tagging and releasing process. However, most of the tagged fish had some mechanical injuries, the majority probably caused by the catching, handling, and tagging processes. The most common injuries were detached scales and fin injuries (Figure 7a,d). Some fish had skin injuries and bruises, probably caused by the abrasion with the netting of a trap (Figure 7a,b). Eye injuries (Figure 7c), injuries caused by meshing, as well cuts and wounds to the body and injuries in the mouth or jaws were observed. The results indicate that injuries are more probable when a salmon is caught with a trap with an emptying chute than when a salmon is caught with a trap with a lifting bag (Figure 8, Table 4). The probability of a lifting bag causing less injuries than a chute was 100% for minor scale loss, 99% for mouth or jaw injuries, 94% for mesh injuries, 90% for eye injuries, and 87% for wounds and scratches. Furthermore, it was more probable for major scale loss or fin injuries to be detected when salmon are caught with a trap with a lifting bag (Table 4).7FIGUREObserved injuries in tagged salmon. (a) Fin and mesh injuries, (b) injuries caused by meshing, (c) eye injury, and (d) major scale loss. Injuries are marked with red arrows.8FIGUREEstimated posterior distributions for probability of a specific type of injury when releasing from a trap. (A) Minor scale loss, (B) major scale loss, (C) eye injury, (D) wounds and scratches, (E) mesh injuries, (F) mouth or jaw injury, and (G) fin injury. Red line: trap with an emptying chute; blue line: trap with a lifting bag.4TABLEEstimated posterior means and 90% highest density probability intervals (HDI) for salmon getting a specific injury when being caught with a pontoon trap with an emptying chute or with a pontoon trap with a lifting bag.Emptying chuteLifting bagInjuryMean90% HDIMean90% HDIMinor scale loss0.67[0.61,0.72]0.51[0.46,0.57]Major scale loss0.18[0.13,0.22]0.41[0.36,0.47]Eye injury0.03[0.01,0.04]0.01[0.00,0.02]Wounds and scratches0.08[0.05,0.11]0.05[0.03,0.7]Mesh injury0.02[0.00,0.03]0.01[0.00,0.01]Mouth or jaw injury0.06[0.03,0.08]0.02[0.00,0.03]Fin injury0.70[0.65,0.75]0.88[0.84,0.92]DISCUSSIONThe Bayesian model developed in this tagging study indicated that the mortality of multi‐sea‐winter salmon released from pontoon traps with an emptying chute was higher than that of those released from pontoon traps equipped with a lifting bag. We designed a rigorous pairwise experiment with the same fishing areas, same fishers, similar handling and tagging process by the same persons, and similar trap design, except for the comparison of the two emptying methods. As external factors were evidently similar for both groups, at least two factors could potentially cause the observed difference in release mortality: (1) the number and quality of injuries caused by the trap and emptying process and (2) the difference in physiological stress caused by the emptying process, for example, in the air exposure time between the methods (e.g., Patterson et al., 2017).Injuries likely caused by fishing gear and/or the emptying process were observed in both emptying method groups. It is difficult to suggest causality between injuries and mortality based on our study design, but in general the probability of an injury was higher in the emptying chute group for most of the injury types. For example, eye injuries were mostly recorded in fish caught with a trap equipped with an emptying chute, as was also found by Östergren et al. (2020) and Ruokonen et al. (2020). In this emptying method, fish jump for a while on a chute before the catch is emptied into a water‐filled tank on a boat. It is possible that this jumping causes internal injuries that cannot be seen by visual inspection and increases delayed mortality after release. On the other hand, the probability of major scale losses and fin injury was higher in the lifting bag group.Air exposure is one of the critical factors during the emptying of fishing gear because it can cause acute mortality or substantial physiological stress that affects survival (Cook et al., 2015). Salmonids are particularly vulnerable to hypoxia, and even short air exposure may reduce viability (Papatheodoulou et al., 2022; Richard et al., 2013). Mortality can occur after a relatively short period (e.g., 1–2 min) in the air (Cook et al., 2015). In our study, air exposure differed between the two emptying methods: air exposure was apparently longer with the traditional emptying chute. In this method, salmon are often exposed to the air for about 30–60 s (or even longer) during the lifting of a fish chamber. With a lifting bag, fish are under the water during the final lifting phase of a trap, and the air exposure time is usually shorter, around 10 s. Patterson et al. (2017) concluded that the extension of handling time in the air to more than 10 s led to increased mortality in Pacific salmon. No specific studies on Atlantic salmon are available, but the difference in air exposure between emptying methods may be one of the main reasons for the difference observed in releasing mortality. Mortality caused by air exposure increases with an increasing water temperature (Patterson et al., 2017; Richard et al., 2013). Water temperatures rose from 7 to 16⁰C during our study, but both groups were tagged following a similar schedule, so it is assumed that this effect was similar for both groups.In our study, release mortality estimates were roughly in the same range as in a previous study conducted in the same area, in which mortality for MSW salmon released from a trap with an emptying chute was estimated to be 25%–31% at most (Ruokonen et al., 2020). However, the Bayesian model used in the present study was more detailed than that used in the previous study because it accounted for the weekly numbers of tagged salmon available to the fisheries, weekly recaptures, and the probability distribution of all components affecting the survival estimates, including the reporting rates of recovered tags at coastal/river fisheries and the tag loss rate, which were treated as fixed quantities in the study of Ruokonen et al. (2020). Siira et al. (2006) estimated maximum release mortality at 20% for MSW salmon released from a traditional pound net used earlier in a coastal fishery on the Bothnian Sea. In both recent studies by our group (Ruokonen et al., 2020), mortality estimates were significantly lower than those found in Swedish studies based on telemetry experiments (Östergren et al., 2020). Their mortality estimates were in the range of 47%–88% with the traditional emptying chute method and between 17% and 63% when a lifting bag was used. However, when comparing the results from different studies, it is important to remember that the tagging and mortality estimation methodologies varied.The release mortalities were estimated using a Bayesian model that utilised informative prior distributions derived from the literature and recent stock assessment results for Baltic salmon (ICES, 2021). This approach enabled us to estimate separately with different fates what tagged salmon may experience during their migration. Those fates included tag loss, escapement to river, and several causes of death. Expert views on tag reporting rates were also elicited and formally averaged into prior distributions. Such a wide use of the available prior information is quite rare in modelling, but we argue that similar approaches would be beneficial for scientific learning and that versatile background information can often be found if its value is recognised and sufficient amount of time and effort is put in the search. Attempts to base the statistical models on biological and systemic realism should also be prioritised over data analytic methods. Methods of statistical data analysis are appropriate when the goal is to describe the observed data with a minimal statistical model. However, when the interest lies in understanding of the system that produced the data, principles of system modelling and inverse inference are more relevant. Bayesian methods show their strength with the former (cf. Kuparinen et al., 2012). The Bayesian system model developed here can be relatively easily applied to a similar study of another fish species, as long as all the components of the model are chosen to be relevant for the case in question.The practical experiences from the present study suggest that the lifting bag type used in our study was unsuitable when the catch was abundant, or when there were large salmon in the catch. When the catch consisted of many large salmon, the opening of the lifting bag often became clogged, which caused fish to press against each other and the netting of the gear, apparently leading to scale loss and fin damage. Furthermore, lifting the bag by hand was quite difficult even when done for smaller proportions of the catch at a time, and during the lifting process, the fish may have suffered additional damage. Many minor injuries were observed in fish hauled by a lifting bag which could expose them to fungal infections later in the spawning river, for example (Weichert et al., 2021). Furthermore, lifting the bag manually is heavy and is a risk to the occupational health of fishers. Emptying the trap with a lifting bag by hand is a major safety risk, an important observation that was also raised by Östergren et al. (2020).One possibility for reducing release mortality could be the improvement of the traditional emptying chute so that it causes less damage to salmon and minimises air exposure. However, it is important to realise that the total effect of release mortality on the salmon population is related to the actual number of fish released. As Östergren et al. (2020) state, no accurate estimate of the number of released salmon is available in Baltic Sea salmon fishery, so an exact numerical estimate of absolute mortality is impossible to produce. According to the current Finnish national fishing regulations for commercial salmon fisheries using traps, the whole catch must be taken and landed unselectively, except for salmon damaged by seals or cormorants, or after the salmon quota per fisher has been fulfilled. The total mortality due to the selection and release of salmon from coastal trap net fishing in Finland is probably low because only a small minority of the total salmon catch is released. In the current situation, the release mortality is not a major issue. Yet, if the fishing rules change in the future to require coastal trap net fishers to release all wild salmon or all large wild salmon, for example, release mortality will become an important issue.AUTHOR CONTRIBUTIONSTimo J. Ruokonen: Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Project administration, Writing original draft, Writing review & editing. Henni Pulkkinen: Data curation, Formal analysis, Methodology, Writing‐review & editing. Samu Mäntyniemi: Conceptualization, Data curation, Formal analysis, Writing‐review & editing. Jaakko Erkinaro: Conceptualization, Methodology, Writing‐review & editing. Petri Suuronen: Conceptualization, Funding acquisition, Investigation, Methodology, Supervision, Writing‐review & editing.ACKNOWLEDGEMENTSWe thank the fishers who participated in this study for their valuable work and assistance during fish tagging. Johan Niskanen and Marko Salminen are thanked for their assistance with the salmon tagging. This study was funded by the Ministry of Agriculture and Forestry of Finland.ETHICS STATEMENTThis research was conducted in accordance with animal ethics approval under FinnishAnimal Welfare Act 1996 (1996/247), which ensures the humane treatment of animals for research.CONFLICT OF INTEREST STATEMENTThe authors declare no conflict of interest.DATA AVAILABILITY STATEMENTThe data that support the findings of this study are available from the corresponding author upon reasonable request.PEER REVIEWThe peer review history for this article is available at https://publons.com/publon/10.1002/aff2.115.REFERENCESArlinghaus, R., Cooke, S.J., Lyman, J., Policansky, D., Schwab, A., Suski, C. et al. (2007) Understanding the complexity of catch‐and‐release in recreational fishing: an integrative synthesis of global knowledge from historical, ethical, social, and biological perspectives. Reviews in Fishery Science, 15, 5–167.Breen, M. & Catchpole, T. (Eds.) (2021) ICES guidelines for estimating discard survival. ICES Cooperative Research. Available at: https://doi.org/10.17895/ices.pub.8006 [Accessed 12 May 2023].Cook, K.V., Lennox, R.J., Hinch S.G. & Cooke, S.J. (2015) Fish out of water: how much air is too much? Fisheries, 40(9), 452–461.EU. (2021) Commission Delegated Regulation (EU) 2021/1417 of 22 June 2021 supplementing Regulation (EU) 2016/1139 concerning the specifications for the landing obligation as regards salmon in the Baltic Sea for the period 2021–2023. Official Journal of European Union, 305, 3–5.Hemmingsson, M., Fjälling, A. & Lunneryd, S. (2008) The pontoon trap: description and function of a seal‐safe trap‐net. Fisheries Research, 93, 357–359.Hoeting, J.A., Madigan, D., Raftery, A.E., & Volinsky, C.T. (1999) Bayesian model averaging: a tutorial (with comments by M. Clyde, David Draper and EI George, and a rejoinder by the authors. Statistical science, 14(4), 382–417.ICES. (2021) Baltic Salmon and Trout Assessment Working Group (WGBAST). ICES Scientific Reports, 3(26), 331.Kuparinen, A., Mäntyniemi, S., Hutchings, J. A. & Kuikka, S. (2012) Increasing biological realism of fisheries stock assessment: towards hierarchical Bayesian methods. Environmental Reviews, 20(2), 135–151.Lennox, R.L., Cooke, S.J., Davis, C.R., Gargan, P., Hawkins, L.A., Havn, T.B. et al. (2017) Pan‐Holarctic assessment of post‐release mortality of angled Atlantic salmon Salmo salar. Biological Conservation, 209, 150–158.Michielsens, C.G.J., McAllister, M.K., Kuikka, S., Mäntyniemi, S., Romakkaniemi, A., Pakarinen, T. et al. (2008) Combining multiple Bayesian data analyses in a sequential framework for quantitative fisheries stock assessment. Canadian Journal of Fisheries and Aquatic Sciences, 65, 962–974.Östergren, J., Blomqvist, C., Dannewitz, J., Palm, S. & Fjälling, A. (2020) Discard mortality of salmon caught in different gears. SLU Aqua , 21Papatheodoulou, M., Závorka, L., Koeck, B., Metcalfe, N.B. & Killen, S.S. (2022) Simulated pre‐spawning catch and release of wild Atlantic salmon (Salmo salar) results in faster fungal spread and opposing effects on female and male proxies of fecundity. Canadian Journal of Fisheries and Aquatic Sciences, 79(2), 267–276.Patterson, D.A., Robinson, K.A., Lennox, R.J., Nettles, T.L., Donaldson, L.A., Eliason, E.J., et al., (2017) Review and evaluation of fishing‐related incidental mortality for Pacific salmon. Available at:https://waves‐vagues.dfo‐mpo.gc.ca/library‐bibliotheque/40569020.pdfRichard, A., Dionne, M., Wang, J. & Bernatchez, L. (2013) Does catch and release affect the mating system and individual reproductive success of wild Atlantic salmon (Salmo salar L.)? Molecular Ecology, 22(1), 187‐200.Rivinoja, P., McKinnel, S. & Lundqvist, H. (2001) Hindrances to upstream migration of Atlantic salmon (Salmo salar) in a northern Swedish river caused by a hydroelectric power‐station. Regulated Rivers‐Research & Management, 17, 101–115.Ruokonen, T.J., Suuronen, P., Pulkkinen, H. & Erkinaro J. (2020) Release mortality of wild Atlantic salmon in coastal pontoon‐trap fishery in the northern Baltic Sea. Fisheries Research, 252, 106336Scientific, Technical and Economic Committee for Fisheries (STECF). (2020) Evaluation of joint recommendations on the landing obligation and on the technical measures regulation. Luxembourg: Publications Office of the European Union.Siira, A., Suuronen, P., Ikonen, E. & Erkinaro, J. (2006) Survival of Atlantic salmon captured in and released from a commercial trap‐net: potential for selective harvesting of stocked salmon. Fisheries Research, 80, 280–294.Suuronen, P., Chopin, F., Glass, C., Løkkeborg, S., Matsushita, Y., Queirolo, D. et al. (2012) Low impact and fuel efficient fishing—looking beyond the horizon. Fisheries Research, 119–120, 135–146.Uhlmann, S. S., & Broadhurst, M. K. (2015) Mitigating unaccounted fishing mortality from gillnets and traps. Fish and Fisheries, 16, 183–229.Weichert, F. G., Axén, C., Förlin, L., Inostroza, P. A., Kammann, U., Welling, A. et al. (2021) A multi‐biomarker study on Atlantic salmon (Salmo salar L.) affected by the emerging red skin disease in the Baltic Sea. Journal of Fish Diseases, 44, 429–440.AAPPENDIX: JAGS MODEL CODE FOR BAYESIAN RELEASE MORTALITY MODELmodel{for(t in 1:T){for(j in 1:2){ # Gear (1: pontoon trap with an emptying chute, 2: PU trap with a lifting bag)N_alive[t,j]∼dbin(p_alive[j],N_tagged[t,j])for(i in 1:2){ # Area (1: sea, 2: river)N_recap[t,i,j]∼dpois(HR[i]*n[t,i,j]*r_rep[i]+0.001)}}}for(j in 1:2){p_alive[j]=exp(‐(h[j]+ip_mark)) # probability of surviving from tagging and release mortalitiesp_die[j]=1‐exp(‐(h[j]+ip_mark*q_hand)) # probability of dying because of releasingn[1,1,j]=N_alive[1,j]n[1,2,j]=0 # no releases in the river# Number of fish at sea over timefor(t in 1:(T‐1)){n[t+1,1,j]=n[t,1,j]*q_SS[1]+N_alive[t+1,j]}# Number of fish in river over timefor(t in 1:T){n[t+1,2,j]=n[t,1,j]*q_SM+n[t,2,j]*q_SS[2]}}# Surviving fractionsq_SS[1]=p_surv[1]*p_keep*(1‐p_move) # survives and stays at seaq_SS[2]=p_surv[2]*p_keep # survives and stays in riverq_SM = p_surv[1]*p_keep*p_move # survives and moves from sea to riverfor(i in 1:2){p_surv[i]=exp(‐(M[i]+F[i])) # survival from natural and fishing mortalities for one weekHR[i]=p_keep*(1‐p_surv[i])*F[i]/(M[i]+F[i]) # harvest rate}# Uninformative priorsp_move∼dunif(0,1) # probability of moving from sea to river, 1 weekfor(j in 1:2){h[j]∼dunif(0,12) # trap‐induced mortality}# Informative priors based on Baltic salmon stock assessment modelF_sea∼dlnorm(log(0.1566)−0.5/T_Fs, T_Fs) # sea F, 11 weeksT_Fs = 1/log(pow(0.02062/0.1566,2)+1)F_river∼dlnorm(log(0.2119)−0.5/T_Fr, T_Fr) # river F, 12 weeksT_Fr = 1/log(pow(0.07043/0.2119,2)+1)M_river∼dlnorm(log(0.0975)−0.5/T_Mr, T_Mr) # river M, 52 weeksT_Mr = 1/log(pow(0.0312/0.0975,2)+1)M_sea∼dlnorm(log(0.1622)−0.5/T_Ms, T_Ms) # sea M, 8 weeksT_Ms ← 1/log(pow(0.052/0.1622,2)+1)# Instantaneous mortalities per weekM[1]=M_sea/8M[2]=M_river/52F[1]=F_sea/11F[2]=F_river/12# Other informative priorsp_mark∼dbeta(1.8,7.2) # mortality due to tagging + handling, 12 weeksip_mark = ‐log(1‐p_mark/12) # instantaneous tagging + handling mortality, 1 week# Share of handling mortality in total tagging + handlingq_hand∼dlnorm(−0.723,7.2485)T(,1)p_lose∼dbeta(12,68) # probability of losing a tag, 12 weeksip_keep = ‐log(1‐loose_tag/12) # probability of keeping tag for one week, instantaneous scalep_keep = exp(‐keep_tag_inst) # probability of keeping a tag for 1 week# Priors for reporting rates of tags based on expert elicitationr_rep[1]=E[Y,1] # sea fisheriesr_rep[2]=E[Y,2] # river fisheriesY∼dcat(w[1:3])for(i in 1:3){w[i]=1/3 # Equal weights given to each expert}# Expert 1E[1,1]∼dbeta(17.25,7.75)E[1,2]∼dbeta(23.1,6.9)# Expert 2E[2,1]∼dbeta(42.6,17.4)E[2,2]∼dbeta(37.75,12.25)# Expert 3E[3,1]∼dbeta(10.14,2.86)E[3,2]∼dbeta(7.28,5.72)}BAPPENDIX: JAGS MODEL CODE FOR BAYESIAN INJURY MODELmodel{for(i in 1:7){ # injury typefor(j in 1:2){ # gear typex[i,j]∼dbin(p[i,j],N[j])p[i,j]∼dbeta(1,1)}}}
Aquaculture Fish and Fisheries – Wiley
Published: Aug 1, 2023
Keywords: Bayesian inference; coastal; fishing technique; injury; Salmo salar; system models; the Baltic Sea
You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.