Severe thiamine deficiency in Baltic Salmon coincides with low wild recruitment, increasingly so at long migration distances

Résumé fourni par la source

Filesrecruitment_data_submit.csv "year": Year of the recruitment electrofishing sampling occasion "river_name": Name of the river "hflodomr": Catchment name "XYKOORLOK": Unique identifier of the site composed by X and Y in RT90 "salmon0": Density (individuals per 100m2) of Atlantic salmon "Site2Mth_sqrt": Distance to mouth in square root meters thiamine_data_submit.csv "Year": Year for the thiamine status estimation. "Year_plus_one": Year + 1 to shift thiamine status on step forward to analyze how recruitment is associated with thiamine status. "River": Name of the river "M74_incidence": M74 incidence, expressed as the percentage of sampled females in a hatchery whose offspring show M74-symptomes. west_coast_data_submit.csv "year": Year for the electrofishing sampling occasion (recruitment). "river_name": Name of the river "XYKOORLOK": Unique identifier of the site composed by X and Y in RT90 "salmon0": Density (individuals per 100m2) of Atlantic salmon thiamine_recruitment_code_submit.R: R Code to reproduce all analyses and plots in the manuscript using the three files above. README.txt: The variable description above.AbstractThiamine deficiency is a condition implicated as a driver of population declines across continents and taxa. In salmon, symptoms of this deficiency are often observed in compensatory hatcheries, where large parts of the clutches display severe neurological disorders and die (called M74). However, whether hatchery-quantified thiamine deficiency is associated with reduced wild salmon recruitment has rarely been tested. Here we combine two long-term monitoring datasets: (i) 30+ year time series of thiamine status in eight Atlantic salmon (Salmo salar) rivers draining into the Baltic Sea, and (ii) wild recruitment data for each corresponding river estimated by electrofishing. We modelled wild recruitment each year as a function of M74 incidence the year before and further tested whether this association was modified by within-river migration length. We found that severe thiamine deficiency coincided with reduced wild recruitment. Furthermore, longer within-river migration steepened the negative association between thiamine deficiency and recruitment. As a control comparison, we did not find any evidence that thiamine status, as quantified in the Baltic Sea, was associated with salmon recruitment on the Atlantic coast. These results indicate that M74, which is mainly observed and quantified in hatcheries, may reflect the conditions experienced by wild salmon in the rivers.MethodsM74 / Thiamine status dataAnnual monitoring of M74, i.e. the prevalence of thiamine deficiency in salmon, in Sweden is performed in compensatory hatcheries and laboratories. M74 incidence is predominantly expressed as the percentage of sampled females in a hatchery whose offspring show M74-symptomes. Annual estimates of M74 incidence for ten river systems in Sweden was acquired from the Baltic Salmon and Trout Assessment Working Group (31). These were, from north to south, Torneälven, Luleälven, Skellefteälven, Ume-/Vindelälven, Ångermanälven, Indalsälven, Ljungan, Ljusnan, Dalälven, and Mörrumsån (Figure 1).The salmon ascending these rivers spent the previous years in their shared feeding ground, i.e. the Baltic Proper (33-35). Since they cease feeding during their sea and river migration, their thiamine status has been hypothesized to be determined in the feeding ground (15, 17). Hence, the outbreaks of M74 and their severity are highly correlated among rivers and fluctuate from over 90% to less than 5% of females producing thiamine deficient offspring (Figure 1) (15, 31). Such offspring show a range of neurological symptoms including “corkscrew” swimming and lethargy followed by a nearly 100% mortality rate (36, 37).Salmon recruitment dataData on juvenile fish density was acquired from the Swedish Electrofishing RegiStry (SERS; https://dvfisk.slu.se/). Electrofishing is a non-lethal sampling method for shallow wade-able rivers, where a direct electric current (DC) is used to attract and catch fish using a hand-held anode and a dip net. Electrofishing is an established and reliable technique for quantifying fish densities in shallow rivers, as detailed by Bohlin et al. (38) and outlined in the European Standard (CEN, 2003; EN 14011:2003). Electrofishing is primarily focused on estimating juvenile (0+ and 1+) abundance and is often aimed specifically at estimating salmonid densities.Only sites where salmon was caught at least once, i.e. likely to be located below the first migration barrier, were included in the dataset, hereafter referred to as salmon sites. Two of the ten river systems had no salmon sites (Luleälven and Ljusnan), probably due to the presence of dams located downstream the first suitable spawning habitat. In the remaining 8 rivers, the number of salmon sites per river varied between 2 and 114 (mean ± SD = 42.9 ± 42.4). The number of (overlapping) years with both M74 and salmon recruitment data varied between 3 and 38 (mean ± SD = 17.4 ± 11.3). The total number of salmon sites in the dataset was n = 250 and the total number of electrofishing occasions (matched with yearly M74 data) was n = 2064.Model specification and selectionThe goal of the statistical model was to assess whether the incidence of M74 in the artificially reared offspring from stripped adult salmon in a given year and river system could explain variation in the wild salmon recruitment the following year and, if so, whether this association varies with migration distance from the river mouth (i.e., energetic demand). To test this, we first constructed a basic model; a generalized linear mixed model (GLMM) with salmon recruitment (individuals per 100 m2) as response variable (at year t), M74 incidence (at year t-1) as explanatory variable and sites nested within rivers as random effects (with varying intercept). We accounted for the temporal auto-correlation of residuals by including an autoregressive term, particularly an Onstein-Uhlenbeck process, which does not assume that sampling years are successive (39). We then built upon this model to check whether adding information would improve parsimony.Response distributionFirstly, the salmon recruitment response variable was positively skewed, as often is the case with density data. To find the best response distribution for the GLMM, we performed the basic model twice, once with a Poisson and once with a negative binomial distribution. To allow analysis with count-based models we transformed the density response variable by multiplying it with 10, which converted it into an integer while preserving relative differences in density (40). Although the transformed variable does not represent a true count variable, it does allow us to use count-based models (40). We found that the negative binomial distribution was a significantly better fit (Poisson, AIC = 19970; negative binomial, AIC = 19636, ΔAIC = 334) and produced residuals with fewer deviations from model assumptions so we used this distribution for all models.Zero-inflationA considerable number of electrofishing occasions had zero-catches (32%), prompting us to test if adding zero-inflation would lead to a more parsimonious model.Migration distanceIt has been suggested that thiamine deficiency can become more pronounced with increased energetic demands such as those associated with long migrations (18, 41, 42). To test whether migration distance modified the effects of thiamine deficiency on recruitment we calculated distance between the mouth and all included study sites in ArcGIS Pro (3.4.1) using the plugin RivEX (1.15) (43) and the 2016 version of the hydrological river network provided by the Swedish Meteorological and Hydrological Institute. We square-root transformed the migration distance variable to reduce positive skew.Random slopes per riverPopulations in different rivers may have different responses of recruitment to M74, prompting us to test whether adding ra

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.