Discussion
This paper documents substantial en route mortality of migrating summer Chum Salmon a major tributary of the Yukon River, the Koyukuk River, in Alaska during the summer of 2019. Although the true numbers of fish that died during this event and the proximate mechanisms of the mortality are not known, available evidence points to a temperature-mediated phenomenon. Comparisons of body size between individuals that successfully accomplished their migrations vs. those that died en route revealed evidence of size-selective mortality favoring larger body sizes, but whether this reflects direct selection acting on size or indirect selection acting on correlated phenotypic traits is unclear. The ultimate question of how this event will impact future summer Chum Salmon production in subsequent years remains to be seen. The lower than average escapement counts observed on the major spawning grounds of Henshaw Creek and the Gisasa River suggests that theen route mortality event was substantial enough to impact the number of spawners surviving at the population level, which raises conservation concerns should die-offs become more frequent and/or greater in severity.
Over a thousand dead fish were counted along a 250 km section of the Koyukuk River during a rapid-response survey on July 26-27, 2019. This work would not have been possible without the invaluable observations of local residents and the use of social media to communicate among communities and with scientists and managers. Local residents have always been the first responders to extreme events, and mechanisms to support communities’ on-going stewardship of the river and its salmon are needed more than ever. The magnitude of the die-off is unknown, but likely substantial given several lines of evidence. First, the survey occurred over a short time period (ca. 1.5 days) and over a relatively small portion of the migratory path. Second, according to local residents and the observed state of carcasses decomposition, it was clear that the survey occurred after the peak of the event, suggesting that counts would have been higher if we had been able to survey the river a few days earlier. Third, carcass counts while traveling were an unknown, but obviously minor, fraction of the number deposited on shore, which in turn is an unknown fraction of the total number that died throughout the drainage. Given the scarcity of data, it is impossible to estimate what the total fraction of the returning fish succumbed toen route mortality. However, observed counts at the two major spawning grounds were below average (Fig. 5), which is attributed by managers to the warm temperatures and observed en route mortality (Brenner et al. 2020). Between 1999 and 2018, Henshaw Creek had averaged 159,000 spawners and between 1994 and 2018 Gisasa River averaged 66,736 spawners. The observations of 34,342 and 19,099 spawners were the fourth smallest on record for both Henshaw Creek and the Gisasa River. It is possible that the anomaly between average and observed escapements was larger for the Henshaw Creek population given the greater energetic demands for populations that must travel further upstream to spawning grounds compared to other populations (Crossin et al. 2004) in addition to the difference in accumulated time in stressful migrating conditions (Hinch et al. 2012).
Although the proximate mechanisms of this en route mortality event are not known, existing evidence is strongly consistent with temperature-mediated effects observed in other locations (Hinch et al. 2012). First, hypoxic conditions may have contributed to the stress of migrants given that measurements of dissolved oxygen indicated approximately 87% saturation. It is important to note that migrating individuals require substantially more oxygen than individuals that are not actively migrating (Brett 1972). Second, summer Chum Salmon returning to the Koyukuk were approximately 8 days later than average (Brenner et al. 2020), which is consistent with the widely observed pattern of delayed migrations in other systems during periods of anomalously warm water. For example, Chinook Salmon in the Klamath River during 2002 entirely ceased migration when the water temperature rose above ca. 20 degrees C, which in turn increased their densities in cool water holding refugia and contributed to the spread of pathogens, particularly Ich and columnaris that ultimately inflicted mass mortality (Strange 2010, 2013). Third, observed water temperatures during the time of the survey averaged 17.1°C, which is above the optimum temperature for migration and is known to put individuals at risk of increased mortality from pathogens such as F. columnare (Holt et al. 1975). Approximately a third of the carcasses were noted to have patches of fungal growth, consistent with earlier infections with columanaris or other bacterial pathogen (Fig. 3), but formal pathology was not conducted. Future rapid response surveys should be prepared to take pathology samples to provide diagnostic disease assessments. Importantly, the individual fish we sampled as carcasses on the mainstem Koyukuk had already traveled 900 km from the mouth of the Yukon River and it is likely that the accumulated time traveling in warm water, combined with delayed migration increased the susceptibility to pathogens, similar to en route mortality events in the Fraser River and Columbia River.
We detected evidence of size-selective mortality, where individuals that died en route were approximately 4% shorter than individuals that survived to the spawning grounds at the Henshaw Creek weir (Fig. 4). This pattern was observed in both males and females and was estimated to be strong selection compared to a global database of standardized selection differentials (Siepielski et al. 2017). However, several important caveats with the analysis are necessary. First, like many analyses of selection this one is plagued by small and unbalanced sample sizes (68 dead vs. 553 alive). Randomization simulations conducted on the combined dataset (alive + dead) revealed that the level of selection observed was likely to occur 1 out of every 4 times due to chance alone. Second, the carcass survey occurred over a much narrower range of dates than the observations at the weir, which may bias results particularly because the size and age of individuals returning to spawning grounds often covaries with run timing (Quinn et al. 2015). Indeed, the size of returning individuals to the Henshaw Weir declined significantly throughout the run (p<0.001), but date explained little of the variance (6%) in size. However, analyses using a date-truncated dataset from the Henshaw weir revealed similar patterns of size selection with large fish being favored to return, and thus the pattern was consistent while including the potential effect of run timing. Third, it is not clear whether size selectivity is the result of direct selection acting on body size or is the result of both direct and indirect selection acting on correlated phenotypic traits such as physiological tolerances and others that comprise a migratory phenotype. The selection analysis here assumes only direct selection, which is likely not true. Future work to survey carcasses throughout the run and to gather additional phenotypic measurements would be needed to formally calculate selection gradients that allow quantification of direct and indirect influences on selection.
Will the en route mortality that occurred in 2019 have a lasting consequence on the populations in the Koyukuk River and how might populations adapt to warming? While this event gives cause for concern given it appears to have resulted in substantially below average spawner escapements, there are also reasons for optimism. Pacific salmon are remarkably productive and resilient in the face of disturbance (Hilborn et al. 2003) and there is little evidence that small spawner numbers can result in positive density-density dependent effects, also known as Allee Effects or depensation (Cunningham et al. 2013). The variable age of adult migrants contributing to any one return year (e.g. fish that spent 2, 3, or 4 winters at sea) provides a natural buffer that will likely dampen (or mask) the true effect of the die-off in future years (Schindler et al. 2010). In contrast, species such as pink salmon (O. gorbuscha ) that all mature at two years of age, might be particularly vulnerable to mortality events. Consistent with observations from other systems in the southern part of the Pacific salmon range, it seems the most obvious response to this and potential future events are shifts in run timing by adults to avoid migrations during stressful periods (Quinn and Adams 1996). Indeed, run timing and migratory behavior, which reflect both plastic and genetic components, are predicted to shift earlier in many systems (Crozier et al. 2011; Reed et al. 2011a). Of course the extent to which run timing may shift is a function of the strength of selection and heritability in addition to plasticity (Carlson and Seamons 2008; Reed et al. 2011b). It is important to recall, however, that run timing is in part also shaped by the selection acting on other parts of the life history, such as the timing of spawning given the water temperature during incubation (Brannon et al. 2004). This presents potential conflicts, as selection may favor earlier return timing on the adult life history to avoid stressful water temperatures, whereas warming should select forlater timing given the relationship with temperature, embryo development, and match-mismatch dynamics. This dynamic selection tug-of-war is worthy of additional exploration using eco-evolutionary modeling.
While the short-term or long-term consequences of 2019 remains to be seen and understood, it is increasingly clear that regional warming throughout Alaska and the Arctic is likely to make these events more frequent and of greater severity. Preparations need to be made to avoid, to the extent possible, surprises like this that are on the horizon. This stark reality underscores the need to protect habitats that serve as the template for life history variation that buffers against disturbance (Moore et al. 2014), ensure migrating individuals have natural options to seek thermal refugia in cooler waters during stressful periods (Armstrong et al. 2016), and support local human communities to continue being the first responders and stewards of salmon, without which all of us would be impoverished.