# Ocean predation and mortality of adult Atlantic salmon

## Abstract

Predation and mortality are often difficult to estimate in the ocean, which hampers the management and conservation of marine fishes. We used data from pop-up satellite archival tags to investigate the ocean predation and mortality of adult Atlantic salmon (Salmo salar) released from 12 rivers flowing into the North Atlantic Ocean. Data from 156 tagged fish revealed 22 definite predation events (14%) and 38 undetermined mortalities (24%). Endothermic fish were the most common predators (n = 13), with most of these predation events occurring in the Gulf of St. Lawrence and from the Bay of Biscay to the Irish Shelf. Predation by marine mammals, most likely large deep-diving toothed whales (n = 5), and large ectothermic fish (n = 4) were less frequent. Both the estimated predation rates (ZP) and total mortality rates (ZM) where higher for Atlantic salmon from Canada, Ireland, and Spain (ZP = 0.60–1.32 y−1, ZM = 1.73–3.08 y−1) than from Denmark and Norway (ZP = 0–0.13 y−1, ZM = 0.19–1.03 y−1). This geographical variation in ocean mortality correlates with ongoing population declines, which are more profound for southern populations, indicating that low ocean survival of adults may act as an additional stressor to already vulnerable populations.

## Introduction

Predation plays a crucial role in structuring aquatic ecosystems by altering the behavior, distribution, and density of prey1,2,3. However, predation is notoriously difficult to quantify because accurate information on prey abundance and prey consumption is required. This is particularly true for pelagic marine ecosystems, which cover large areas and are highly dynamic in terms of species composition and abundance4,5.

Improving estimates of predation and natural mortality is of the utmost importance to fisheries assessment and management, particularly for vulnerable populations6,7. A prime example of a species for which detailed information about predation and mortality is needed is the anadromous Atlantic salmon (Salmo salar). Since the early 1980s, Atlantic salmon have experienced prolonged population declines, partially due to reduced survival during their ocean feeding migration8. While these declines are occurring throughout the species’ distribution range, the negative trend is most profound for the southernmost populations8.

During the marine phase most Atlantic salmon migrate to oceanic feeding areas that are distant from their river of origin, where they are predominantly found in pelagic habitats9,10. Knowledge of their ocean ecology is largely limited to foraging11,12, with most of the information about marine mortality originating from studies carried out close to natal rivers13 (but see14). In the ocean, Atlantic salmon constitute a minor fraction of the prey field and they are rarely documented as prey even for the most important predators13. Thus, in order to obtain quantitative descriptions of the ocean predation and mortality of Atlantic salmon, information needs to be collected from the perspective of the prey to avoid impractically large sample sizes.

Rapid developments in archival telemetry have allowed a much greater understanding of the ocean distribution for numerous fish species15,16,17. In studies of the ocean migration for large pelagic fishes, the most commonly used tag type is the pop-up satellite archival tag (PSAT), which records temperature, depth and light data, and is programmed to detach, surface, and transmit a subset of the archived data to satellites on a pre-determined date10,18. Premature detachment occurs if a constant depth is detected over a multi-day time period or if the tag records an extreme depth endangering its physical integrity. Under such circumstances, inference can be made about whether or not a tagged individual has died19. For obvious predation events, predator species can be inferred by comparing the temperature and depth data with behavioral patterns previously recorded for potential predators20,21.

In contrast to Pacific salmon species (Oncorhynchus spp.), Atlantic salmon are iteroparous. Repeat spawners can be important contributors to recruitment22,23, suggesting that increased predation on these fish could hinder the recovery of vulnerable populations21. Here, the open-ocean predation and mortality of adult Atlantic salmon were investigated during their repeat ocean migration by using PSATs. A total of 227 Atlantic salmon were tagged in 12 rivers in Canada, Denmark, Ireland, Norway, and Spain (Table 1). Our objectives were to 1) quantify the open-ocean mortality rates of the tagged Atlantic salmon; 2) investigate the geographical distribution of predation events; and 3) identify the most likely predators. In our analyses, Atlantic salmon populations from Canada, Denmark, and Ireland were grouped according to their country of origin due to the geographical proximity of the study rivers, while the Norway populations (northeast, northwest, and central) and Spain population were treated as unique groups.

## Results

Marine migration records were obtained from 156 of the 227 deployed tags, with the remaining 71 tags failing to report (Fig. 1). Tags were deployed for 1–313 days, with the longest deployments for the Norwegian groups (Table 1). Of the 156 tags that reported, 32 transmitted their data on the scheduled date, 3 tags were retrieved from consecutive spawners that returned to the river after spending the summer at sea, and 4 fish were recaptured in marine fisheries (Table 1). These 39 fish are known to have remained alive throughout the deployment period, while the remaining 117 tags were classified as premature detachments. Inspections of light, temperature, and depth data from all 117 premature detachments revealed 22 definite predation events, 38 mortality events for undetermined reasons, and 57 premature detachments from live fish for unknown reasons (Table 1). The percentage of reporting tags detaching prematurely from live fish for unknown reasons varied between 5–71% for the different groups and was highest for the Norwegian populations (50–71%).

Predation events occurred 8–159 days after release (median = 39 d, SD = 35 d) and accounted for 14% of the reporting tags. Further investigations of the temperature profiles for the ingested tags revealed that 5 Atlantic salmon were consumed by marine mammals, 4 were consumed by ectothermic fish, and 13 were consumed by endothermic fish (Table 1), with a significant difference in proportions between the predator groups (χ2 test for goodness of fit: χ2df=2, n=22 = 6.64, p-value = 0.04). Predation was inferred for fish from Canada, Denmark, Ireland, and Spain (Fig. 1). Although Norwegian Atlantic salmon accounted for 42% of the reporting tags, no predation events were recorded for these three populations. The proportion of the tagged fish experiencing predation was 0.42 (95% CI = 0.14–0.70) for the Spanish group, 0.36 (95% CI = 0.18–0.53) for the Canadian group, 0.26 (95% CI = 0.07–0.46) for the Irish group, and 0.06 (95% CI = 0–0.15) for the Danish group. By accounting for the mean deployment duration, the overall estimated instantaneous mortality rate (ZP) was 0.29 y−1 for all groups combined. For the groups experiencing predation, ZP was 1.32 y−1 for the Canadian fish, 1.06 y−1 for the Spanish fish, 0.60 y−1 for the Irish fish, and 0.13 y−1 for the Danish fish.

Undetermined mortality events occurred 1–205 days after release (median = 16 d, SD = 49 d). This accounted for 24% of the reporting tags and included 34 tags with records indicating that they transmitted data after a period on the ocean floor, 3 tags that recorded depths exceeding the tag threshold, and 1 tag that was scavenged. Atlantic salmon from all areas experienced undetermined mortalities, primarily in coastal waters off Europe, in the Gulf of St. Lawrence, and in oceanic waters near the Jan Mayen Island (Fig. 1). The proportion of tagged fish that died (predation and undetermined mortality combined) was 0.79 (95% CI = 0.61–0.97) for the Irish group, 0.58 (95% CI = 0.30–0.86) for the Spanish group, 0.54 (95% CI = 0.35–0.72) for the Canadian group, 0.41 (95% CI = 0.24–0.58) for the Danish group, 0.29 (95% CI = 0.05–0.52) for the northeast Norway group, 0.12 (95% CI = 0.02–0.22) for the northwest Norway group, and 0.10 (95% CI = 0–0.29) for the central Norway group. For all groups combined, this corresponded to an estimated total instantaneous mortality rate (ZM) of 0.92 y−1. For the different groups, ZM was 3.08 y−1 for the Irish fish, 2.29 y−1 for the Canadian fish, 1.73 y−1 for the Spanish fish, 1.03 y−1 for the Danish fish, 0.55 y−1 for the northeast Norway fish, 0.20 y−1 for the central Norway fish, and 0.19 y−1 for the northwest Norway fish.

Several of the tagged fish from Canada, Denmark, and northwest Norway remained alive throughout the entire deployment period (Table 1). No body size difference was detected between dead and surviving fish from Canada and northwest Norway (permutation tests: Z = 0.26 and 0.49, p-values = 0.79 and 0.62, respectively). For fish from Denmark, surviving fish were significantly larger than fish that died (permutation test: Z = 2.27, p-value = 0.02).

### Predation by marine mammals

The five Atlantic salmon eaten by marine mammals originated from Canada (n = 1), Denmark (n = 1), and Ireland (n = 3), and accounted for 23% of the predation events. All four tags attached to European Atlantic salmon surfaced in waters south of Iceland, whereas the tag attached to the Canadian Atlantic salmon surfaced in the Labrador Sea (Fig. 2). Between ingestion and expulsion, the maximum depth ranged from 215–818 m (mean = 577 m, SD = 240 m); three of the tags recorded frequent diving to depths below 400 m (Table 2, Figs 2 and S1).

The tag consumed by a marine mammal in the Labrador Sea was successfully retrieved, thus allowing a more detailed investigation. During the 84-h ingestion period, 320 dives were identified. These dives ranged in duration from 4–18 min (mean = 8 min, SD = 4 min) and maximum depths during dives ranged from 10–761 m (mean = 240 m, SD = 270 m). Both the duration and maximum depths of the dives had a bimodal distribution (Fig. 2). The archived data revealed substantial drops in temperature, primarily associated with diving, and the lowest temperature recorded was 27.6 °C (Fig. 2).

### Predation by ectothermic fish

Predation by ectothermic fish occurred across the Atlantic Ocean and predation was recorded for Atlantic salmon released from Canada (n = 1), Denmark (n = 1), Ireland (n = 1), and Spain (n = 1). Ectothermic predation accounted for 18% of the predation events. Three predators displayed a non-surface-oriented behavior, whereas one predator spent 61% of its time in the upper 20 m of the water column (Figs 3, S2). Overall, maximum depths ranged from 172–915 m (mean = 570 m, SD = 339 m) (Table 2).

Pop-up locations for the tags consumed by deep-dwelling ectothermic fish were in the Gulf of St. Lawrence and south of Iceland (Fig. 3). The depth profiles between ingestion and expulsion documented dives over large depth intervals (98–422 m) (Figs 3, S2).

The tag consumed by a pelagic ectothermic fish surfaced north of the Azores (Fig. 3). The vertical profile between ingestion and expulsion showed regular dives to depths exceeding 200 m that were either continuous throughout most of the daylight period or disrupted by short surfacing (i.e. basking) events (Fig. 3). Greater depths were utilized during the day than at night (permutation test: Z = 18.15, p-value = 2.2 × 10−16, Fig. 3). Vertical movements were confined to the isothermal mixed layer, with temperature recordings ranging from 11.5–14.6 °C.

### Predation by endothermic fish

Predation by endothermic fish accounted for 59% of the predation events and was evident for Atlantic salmon released from Canada (n = 8), Ireland (n = 1), and Spain (n = 4).

#### Northwest Atlantic Ocean

Seven of the eight predation events in the Northwest Atlantic occurred in the Gulf of St. Lawrence while one occurred over the Labrador Shelf (Figs 4, S3). The time spent in surface waters (upper 20 m of the water column) varied between 35–99% among the predators (Table 3). The maximum depth ranged from 22–355 m (mean = 158 m, SD = 103 m) (Table 3).

For the Atlantic salmon eaten by predators in the Gulf of St. Lawrence, identification of the most likely predators was carried out with a linear discriminant analysis (LDA), with Atlantic bluefin tuna (Thunnus thynnus) and porbeagle (Lamna nasus) as the candidate predators (see Methods). The LDA successfully separated the two species based on percentage of time spent in the upper 20 m of the water column (% surface), the difference in maximum depth between day and night (diel amplitude), and the 95th percentiles of all recorded depths (95%). Atlantic bluefin tuna spent more time in surface waters compared to the porbeagle, which utilized deeper waters and displayed greater diel amplitudes when in certain behavioral modes (Table 3). In four of the seven predation events, Atlantic bluefin tuna (Thunnus thynnus) were considered the most likely predator, while the remaining three were assigned to porbeagle (probabilities 0.9–1) (Table 3).

The maximum gut temperature (TMax) ranged from 21.9–26.9 °C for the tags assigned as consumed by Atlantic bluefin tuna, and from 26.2–26.9 °C for the tags assigned as consumed by porbeagle (Table 3). For the tag ingested over the Labrador Shelf, TMax was 24.1 °C (Table 3). The difference between the maximum gut temperature and the mean ambient water temperature after expulsion (TE) ranged from 11.5–17.9 °C for tags consumed by Atlantic bluefin tuna and from 13.7–17.4 °C for those consumed by porbeagle (Table 3). For the tag consumed over the Labrador Shelf, TE was 16.3 °C (Table 3). All predators displayed variation in stomach temperatures and the difference between the maximum and the minimum temperature (TD) ranged from 5.5–10.7 °C (Table 3).

#### Northeast Atlantic Ocean

All of the Spanish Atlantic salmon eaten by endothermic fish (n = 4) were consumed between the Irish Shelf and the Bay of Biscay, while the Irish Atlantic salmon (n = 1) was eaten in the Irminger Sea (Figs 4, S4). In contrast to the Atlantic salmon eaten by endothermic fish in the Gulf of St. Lawrence, a quantitative assessment of the most likely predators was not done because suitable data from potential predators were not available. Overall, the endothermic fish spent 49–100% of their time in surface waters (depths < 20 m) and the maximum depth ranged from 11–737 m (mean = 363 m, SD = 289 m) (Table 3). TMax values of predators ranged from 23.6–29.1 °C and TE from 11.8–18.6 °C (Table 3). Variation in stomach temperature was recorded by all tags, with TD ranging from 4.8–9.9 °C (Table 3).

## Discussion

This study is the first to describe ocean predation and mortality of adult Atlantic salmon over large parts of the species’ distribution range and provides direct evidence of predation by both endothermic and ectothermic predators. Marine mammals have previously been recorded feeding on Atlantic salmon24,25, but this is the first study to record mammal predation on PSAT tagged individuals. Between ingestion and expulsion, all five tags consumed by marine mammals recorded depths exceeding 200 m, indicating predation by large deep-diving toothed whales26,27,28. Other marine mammals, such as pinnipeds, orca (Orcinus orca), and other oceanic dolphins (Delphinidae), tend to tear apart large salmonids at the surface24,25,29, which is unlikely to lead to tag ingestion. For the Atlantic salmon consumed in the Northeast Atlantic Ocean, further identification of likely predators was not possible because the transmitted data were either distorted by the tags’ data compression procedure, or there was insufficient temporal resolution of data for comparison with behavioral data from deep-diving whales. In contrast, for the Atlantic salmon consumed in the Labrador Sea, the potential predator could be inferred due to the high-resolution of the data recovered from the tag. Among the deep-diving toothed whales found in the Labrador Sea whose diving behaviour has been studied, the duration and depth of the recorded dives matched that of the long-finned pilot whale (Globicephala melas)26 and the beluga whale (Delphinapterus leucas)27, indicating that the Atlantic salmon consumed in the Labrador Sea was probably eaten by one of these species.

Our study is the first to provide evidence of predation on PSAT tagged adult Atlantic salmon by deep-dwelling endothermic fish. However, little is known about the spatial distribution and vertical movements of large deep-dwelling fish, which makes postulating the most likely predators difficult. In contrast, predation by a pelagic ectothermic fish has previously been observed for PSAT tagged Atlantic salmon21. For the individual consumed by a pelagic ectotherm north of the Azores in the present study, the tag recorded a series of long-lasting daytime dives within the isothermal mixed layer, occasionally disrupted by short basking events. This pattern resembles the vertical behavior recorded for blue sharks (Prionace glauca)30,31 and swordfish (Xiphias gladius)32,33, indicating that one of these species was the likely predator.

Endothermic fish were more often predators of Atlantic salmon than other predator groups. This corroborates previous observations of marine predation on adult Atlantic salmon21, indicating that populations co-occurring with large endothermic fish may be particularly vulnerable to adult predation. Of the four species of endothermic fish recorded in the Gulf of St. Lawrence, Atlantic bluefin tuna and porbeagle are by far the most common34,35, while records of white shark (Carcharadon carcharias) and shortfin mako (Isurus oxyrinchus) are largely limited to infrequent catches and rare sightings36,37. The predation on Atlantic salmon by endothermic fish in the Gulf of St. Lawrence overlapped temporally with the presence of large-sized Atlantic bluefin tuna and porbeagle34,35 and we consider it unlikely that these predation events were caused by other endothermic fish. In a previous study of North American Atlantic salmon tagged with PSATs, predation by lamnid sharks (Lamnidae) and Atlantic bluefin tuna was spatially segregated, with porbeagle being the most likely predators in the Bay of Fundy and Atlantic bluefin tuna over the Scotian Shelf21. In our study, no clear spatial divergence was detected, indicating that predation by Atlantic bluefin tuna and porbeagle overlapped within the Gulf of St. Lawrence.

For the remaining Atlantic salmon eaten by endothermic fish, the vertical movements recorded by most of the tags consumed adjacent to the Bay of Biscay and the Irish Shelf resembled a behavior commonly seen in porbeagle, with a profound diel pattern in depth use and long-lasting deep dives38,39. This suggests predation by porbeagle, which is supported by a spatial-temporal overlap between the predation events and the distribution of porbeagle in the Northeast Atlantic Ocean40. However, as no quantitative analysis was performed, this conclusion should be treated with caution, and predation by other large endothermic fish, such as the shortfin mako and Atlantic bluefin tuna, cannot be excluded. For the Atlantic salmon consumed over the Labrador Shelf and in the Irminger Sea, determining the most likely predator was not possible, because little is known about the horizontal distribution and vertical movements of endothermic fish in these waters.

While the observed predation on Atlantic salmon explicitly reveal novel ecological interactions, the total mortality imposed on the different groups is perhaps of even greater importance from a management perspective. It is possible that some of the undetermined mortalities were due to predation by smaller marine animals that are unlikely to completely ingest an adult Atlantic salmon either due to gape-size limitations or feeding tactics. In particular, some of the undetermined mortalities observed in coastal areas with dense seal populations could be from seal predation41,42, which has been suggested to have a significant impact on certain Atlantic salmon populations24. The recent population recoveries of marine mammals, including seal herds in the Gulf of St. Lawrence, are considered conservation success stories41. However, these rehabilitations may impose new challenges when both predator and prey are threatened43.

Most efforts to quantify the ocean mortality of Atlantic salmon have focused on the survival of first time migrants, for which the annual mortality varies from 70–99%, both temporally among years and spatially between rivers8. However, estimates of adult mortality at sea exist for some populations. Based on adult Atlantic salmon tagged with acoustic tags in northern Norway, an estimated ocean survival corresponding to an instantaneous mortality rate of 1.17 y−1 was recorded44. In a North American river, the mean instantaneous mortality rates of adult Atlantic salmon for three 10-year periods ranged from 0.60–0.74 y−1, with some annual mortality rates exceeding 5.0 y−1, reflecting an absence of repeat spawners in the reproducing stock22. When comparing our results with these studies, it is possible that our total instantaneous mortality rate may have been underestimated for the Norwegian population, which experienced lower mortality compared to the other groups and no confirmed predation. Notably, a greater proportion of tags detaching prematurely for unknown reasons was observed for the Norwegian populations, and some of these events could be due to mortality. However, even if every premature detachment was caused by mortality, the combined mortality of the Norwegian Atlantic salmon would still not exceed that of Atlantic salmon from the other groups. This suggests that the observed geographical trend in ocean mortality reflects a genuine variation in mortality regimes imposed on the tagged fish, even though the estimated predation and total mortality rates are somewhat uncertain. To what extent these total mortality rates influence trends in the population abundance of Atlantic salmon is uncertain. However, as the spatial pattern present in our data correlates with the ongoing trends in population abundance8, it is possible that low survival of adult Atlantic salmon may act as an additional stressor to already vulnerable populations.

The use of animal telemetry data for quantifying marine mortality has received increased attention in recent years, particularly with the design of acoustic receiver networks that can provide direct observations of mortality in certain systems45. However, spatially independent information about how, where, and when an individual died while in the open ocean is currently only feasible using PSATs. Despite the increasing evidence of novel interactions between predators and PSAT tagged prey, the use of PSATs for accurately describing mortality may in some cases be difficult, as tagging effects may distort the estimates46. In this study, mortality and predation rates were higher for the populations comprised by smaller sized individuals, and our result may be impacted by a greater physical impediment imposed by the tags on these fish. However, as no size difference was detected regarding the fate of the tagged fish in two of the three groups with several confirmed mortalities and several fish alive until tag detachment, we argue that the geographical differences in predation and mortality rates are at least partly representative of contrasting mortality regimes.

In conclusion, ocean predation and mortality of Atlantic salmon varied largely among geographical areas. Estimated predation and total mortality rates were low for Atlantic salmon from northern Europe, with no confirmed predation of fish originating from Norway. This contrasted to the Atlantic salmon released from Canada, Spain, and Ireland, where higher predation and total mortality rates were estimated. The observed predator diversity demonstrated that a variety of large aquatic animals might forage opportunistically on Atlantic salmon during their ocean migration, with a particularly high predation from large endothermic fish.

## Materials and Methods

### Tagging procedure

In the period 2008–2016, a total of 227 adult Atlantic salmon were tagged with pop-up satellite archival tags (PSATs) in several rivers of the North Atlantic basin (Table 1). The length of the tagged fish ranged from 63–115 cm, and the mean length of the different groups ranged from 73 cm (Ireland) to 99 cm (northwest Norway) (Table 1). All Atlantic salmon were tagged in the spring before migrating to sea (March 11 – May 31), with date of tagging depending on location (Table 1). PSATs were deployed externally by attaching the tag to two cushioned back plates that were wired through the dorsal musculature of the fish below the dorsal fin9,10. Fish tagging was approved by the Animal Experimentation Council Denmark, the Department of Fisheries and Oceans Canada, the Department of Health and Children Ireland, the Norwegian Animal Research Authority, and the Ponteverda Province Spain in accordance with national laws for experiments using live animals.

### Tag details

The PSATs (X-tag, Microwave Telemetry, Inc.) were 120 mm long (273 mm including the antenna), had a diameter of 32 mm, weighed 40 g in air, were slightly buoyant, and had a lifespan of 16 months. Tags were programmed to release after 86–313 days, if the pressure was constant for 3–5 days, or if the pressure sensor recorded depths endangering its physical integrity (manufacturer specified at 1250 m). During deployment, X-tags recorded temperature, depth, and light intensity at two-minute intervals. The complete data set was only available if tags were retrieved and the transmitted data only contained a subset of the archived information because of limited tag battery life and bandwidth of the Argos satellite system. The X-tag implemented data compression techniques prior to transmission, which could cause tags to report distorted temperature and depth values. These values were present if the rate of change exceeded a certain threshold. This causes overestimation of variables during rapid decrease and underestimation of variables during rapid increase (http://www.microwavetelemetry.com).

### Quantifying predation events and mortalities

Premature tag detachment occurred due to predation, undetermined mortality, and for unknown reasons. For tags detaching prematurely, tag data were investigated to identify the fate of the tagged fish. Tag consumption was recognized by the tag failing to record light during periods when daylight should have been detected, in combination with observations of vertical movements. For these tags, the vertical profile around the period of mortality was scrutinized to identify whether the tagged Atlantic salmon was eaten alive or scavenged after its death (Supplementary Fig. S5). Scavenging was inferred if the tag initially recorded a steady descent through the water column, indicating that the Atlantic salmon had died before the tag was ingested (Supplementary Fig. S5). Predation was inferred when this pattern was absent (Supplementary Fig. S5). For Atlantic salmon eaten by predators, temperature recordings were further investigated (Supplementary Fig. S5). If the temperature increased above the ambient water temperature, predation by an endotherm was inferred. Tags that recorded stomach temperatures of 37 °C were classified as consumed by marine mammals, while tags that recorded stomach temperatures above ambient but well below 37 °C were classified as consumed by endothermic fish because visceral temperatures in warm-gutted fish do not reach this threshold47,48. If there was no abrupt increase in temperature after ingestion, predation by an ectotherm was inferred.

For tags reporting prematurely with no evidence of ingestion, tag data were inspected to determine if there were clear signs that the fish had died or not. Undetermined fish mortality was inferred if the tag was scavenged, if it transmitted data after recording constant pressure at depths, or if it recorded a steady descent before triggering the pressure fail-safe mechanism. The reasoning behind this is that the slight buoyancy of the X-tag, implies that sinking tags and tags that were assumed lying on the ocean floor were attached to parts of an Atlantic salmon carcass (Supplementary Fig. S5). Tags that transmitted data after recording constant pressure at the surface were not indicative of fish mortality and characterized to detach prematurely for unknown reasons. To obtain a conservative measurement of predation and total mortality rates, these fish were considered alive until tag detachment, because they could be live fish that had lost their tag either due to attachment failure or due to the fish cruising at the surface for a prolonged period, triggering the tag’s release mechanism. The total mortality in each group included fish experiencing undetermined mortality and predation. Instantaneous predation (ZP) and total mortality (ZM) rates were calculated following Ricker 197549:

$${{\rm{Z}}}_{{\rm{P}}}=[-\,\,\mathrm{ln}({{\rm{N}}}_{{\rm{P}}}{/{\rm{N}}}_{1})]/{\rm{t}}$$
$${{\rm{Z}}}_{{\rm{M}}}=[-\,\,\mathrm{ln}({{\rm{N}}}_{{\rm{M}}}{/{\rm{N}}}_{1})]/{\rm{t}}$$

where NP/N1 is the proportion of the sample that did not experience predation, NM/N1 is the proportion of the sample not experiencing undetermined mortality or predation, and t is the mean deployment duration given as the fraction of one year. Based on these rates, exponential survival functions were formalized for each group:

$${{\rm{S}}}_{{\rm{\Pr }}}={{\rm{e}}}^{-{\rm{ZT}}}$$

where SPr is the survival probability, Z is the instantaneous rate, and T is time given as the fraction of one year. The difference in body length between fish that died and fish that remained alive throughout the deployment period was investigated independently for each group using permutation tests. Only groups with several fish alive throughout the deployment period and several confirmed mortalities were tested.

### Predator identification

The method of predator identification varied depending on predator type and geographical location of the predation events. For tags ingested by marine mammals, ectotherms, and endothermic fish in the Northeast Atlantic Ocean, predator identification was limited to comparing behavioral data from candidate predators against data recorded by the tags.

For Atlantic salmon consumed by endothermic fish in the Northwest Atlantic Ocean, predation by either Atlantic bluefin tuna or porbeagle was assumed a priori, because these are the warm-gutted fish most frequently occupying waters overlapping the observed predation events34,35. The most likely predators were identified by a linear discriminant analysis (LDA) because data on the vertical behavior of both Atlantic bluefin tuna and porbeagle were available from waters in proximity to the predation events19,38. The LDA identifies the linear combination of variables that creates the greatest between-group variance for objects with known affiliation, which in a subsequent step can be used to predict the affinity of unknown entities20.

LDA variables were derived from the vertical time series of four Atlantic bluefin tuna tagged in the Gulf of St. Lawrence in autumn19 and from two porbeagle tagged off eastern Canada during summer38 (Table 3, Supplementary Fig. S6). These reference data had a higher temporal resolution than the data from the consumed tags because they were obtained from physically recovered tags (Table 3). The reference time series were therefore sub sampled to match the resolution of the consumed tags. For the Atlantic bluefin tuna, only data from the period when the fish were resident in the Gulf of St. Lawrence were included in the reference data set. For the porbeagle, data until November 1 were used. To correct for post-release behavior50, the initial period following tagging was investigated for all reference tags. If a modified behavior was detected shortly after tagging, days including this behavior were omitted from the analysis. Furthermore, as both porbeagle used for predator reference displayed distinct vertical movement patterns, a split-moving window analysis was conducted in order to objectively separate different behavioral modes51 (see Supplementary Information for details). Distinct behavioral modes were treated as independent entities in the LDA (Table 3).

## Data Availability

Data will be made available upon request.

## References

1. 1.

Baum, J. K. & Worm, B. Cascading top-down effects of changing oceanic predator abundances. J. Anim. Ecol. 78, 699–714 (2009).

2. 2.

Romare, P. & Hansson, L. A. A behavioral cascade: top-predator induced behavioral shifts in planktivorous fish and zooplankton. Limnol. Oceanogr. 48, 1956–1964 (2003).

3. 3.

Swain, D. P., Benoît, H. P. & Hammill, M. O. Spatial distribution of fishes in a Northwest Atlantic ecosystem in relation to risk of predation by a marine mammal. J. Anim. Ecol. 84, 1286–1298 (2015).

4. 4.

Block, B. A. et al. Tracking apex marine predator movements in a dynamic ocean. Nature 475, 86–90 (2011).

5. 5.

Sequeira, A. M. M. et al. Convergence of marine megafauna movement patterns in coastal and open oceans. Proc. Natl. Acad. Sci. 115, 3072–3077 (2018).

6. 6.

Hammill, M. O., Stenson, G. B., Swain, D. P. & Benoît, H. P. Feeding by grey seals on endangered stocks of Atlantic cod and white hake. ICES J. Mar. Sci. 71, 1332–1341 (2014).

7. 7.

Chasco, B. E. et al. Competing tradeoffs between increasing marine mammal predation and fisheries harvest of Chinook salmon. Sci. Rep. 7, 15439, https://doi.org/10.1038/s41598-017-14984-8 (2017).

8. 8.

ICES. Report of the Working Group on North Atlantic Salmon (WGNAS). ICES CM 2018/ACOM:21 (2018).

9. 9.

Hedger, R. D. et al. Diving behaviour of Atlantic salmon at sea: effects of light regimes and temperature stratification. Mar. Ecol. Prog. Ser. 574, 127–140 (2017).

10. 10.

Strøm, J. F. et al. Ocean migration of pop-up satellite tagged Atlantic salmon from the Miramichi River in Canada. ICES J. Mar. Sci. 74, 1356–1370 (2017).

11. 11.

Rikardsen, A. H. & Dempson, J. B. Dietery life-support: The food and feeding of Atlantic salmon at sea in Atlantic Salmon Ecology (eds Aas, Ø., Einum, S., Klemetsen, A. & Skurdal, J.) 115–143 (Blackwell Publishing Professional, 2011).

12. 12.

Dixon, H. J., Dempson, J. B., Sheehan, T. F., Renkawitz, M. D. & Power, M. Assessing the diet of North American Atlantic salmon (Salmo salar L.) off the West Greenland coast using gut content and stable isotope analyses. Fish. Oceanogr. 26, 555–568 (2017).

13. 13.

Cairns, D. K. A review of predator-prey and competitive inter-specific interactions in Atlantic Salmon (Salmo salar). DFO Canadian Science Advisory Secretariat 2006/019 (2006).

14. 14.

Montevecchi, W. A., Cairns, D. K. & Myers, R. A. Predation on marine-phase Atlantic salmon (Salmo salar) by gannets (Morus bassanus) in the Northwest Atlantic. Can. J. Fish. Aquat. Sci. 59, 602–612 (2002).

15. 15.

Block, B. A. et al. Electronic tagging and population structure of Atlantic bluefin tuna. Nature 434, 1121–1127 (2005).

16. 16.

Righton, D. et al. Empirical observations of the spawning migration of European eels: the long and dangerous road to the Sargasso Sea. Sci. Adv. 2, e1501694, https://doi.org/10.1126/sciadv.1501694 (2016).

17. 17.

Strøm, J. F., Thorstad, E. B., Hedger, R. D. & Rikardsen, A. H. Revealing the full ocean migration of individual Atlantic salmon. Anim. Biotelemetry 6, 2, https://doi.org/10.1186/s40317-018-0146-2 (2018).

18. 18.

Aarestrup, K. et al. Oceanic spawning migration of the European eel (Anguilla anguilla). Science. 325, 1660 (2009).

19. 19.

Stokesbury, M. J. W., Neilson, J. D., Susko, E. & Cooke, S. J. Estimating mortality of Atlantic bluefin tuna (Thunnus thynnus) in an experimental recreational catch-and-release fishery. Biol. Conserv. 144, 2684–2691 (2011).

20. 20.

Béguer-Pon, M. et al. Shark predation on migrating adult American eels (Anguilla rostrata) in the Gulf of St. Lawrence. PLoS One 7, e46830, https://doi.org/10.1371/journal.pone.0046830 (2012).

21. 21.

Lacroix, G. L. Large pelagic predators could jeopardize the recovery of endangered Atlantic salmon. Can. J. Fish. Aquat. Sci. 71, 343–350 (2014).

22. 22.

Hubley, P. B. & Gibson, A. J. F. A model for estimating mortality of Atlantic salmon, Salmo salar, between spawning events. Can. J. Fish. Aquat. Sci. 68, 1635–1650 (2011).

23. 23.

Chaput, G. & Jones, R. Reproductive rates and rebuilding potential for two multi-sea-winter Atlantic salmon (Salmo salar L.) stocks of the Maritime provinces. DFO Canadian Science Advisory Secretariat 2006/027 (2006).

24. 24.

Middlemas, S. J., Armstrong, J. D. & Thompson, P. M. The significance of marine mammal predation on salmon and sea trout in Salmon at the Edge (ed. Mills, D.) 43–60 (Blackwell Science, 2003).

25. 25.

Vester, H. & Hammerschmidt, K. First record of killer whales (Orcinus orca) feeding on Atlantic salmon (Salmo salar) in northern Norway suggest a multi-prey feeding type. Mar. Biodivers. Rec. 6, e9, https://doi.org/10.1017/S1755267212001030 (2013).

26. 26.

Heide-Jørgensen, M. P. et al. Diving behaviour of long-finned pilot whales Globicephala melas around the Faroe Islands. Wildlife Biol. 8, 307–313 (2002).

27. 27.

Citta, J. J., Suydam, R. S., Quakenbush, L. T., Frost, K. J. & O’Corry-Crowe, G. M. Dive behavior of Eastern Chukchi beluga whales (Delphinapterus leucas), 1998–2008. Arctic 66, 389–406 (2013).

28. 28.

Watwood, S. L., Miller, P. J. O., Johnson, M., Madsen, P. T. & Tyack, P. L. Deep-diving foraging behaviour of sperm whales (Physeter macrocephalus). J. Anim. Ecol. 75, 814–825 (2006).

29. 29.

Roffe, T. J. & Mate, B. R. Abundance and feeding habits of pinnipeds in Rouge River, Oregon. J. Wildl. Manage. 48, 1262–1274 (1984).

30. 30.

Campana, S. E. et al. Migration pathways, behavioural thermoregulation and overwintering grounds of blue sharks in the Northwest Atlantic. PLoS One 6, e16854, https://doi.org/10.1371/journal.pone.0016854 (2011).

31. 31.

Howey, L. A., Wetherbee, B. M., Tolentino, E. R. & Shivji, M. S. Biogeophysical and physiological processes drive movement patterns in a marine predator. Mov. Ecol. 5, 16, https://doi.org/10.1186/s40462-017-0107-z (2017).

32. 32.

Dewar, H. et al. Movements and behaviors of swordfish in the Atlantic and Pacific Oceans examined using pop-up satellite archival tags. Fish. Oceanogr. 20, 219–241 (2011).

33. 33.

Sepulveda, C. A., Knight, A., Nasby-Lucas, N. & Domeier, M. L. Fine-scale movements of the swordfish Xiphias gladius in the Southern California Bight. Fish. Oceanogr. 19, 279–289 (2010).

34. 34.

Hanke, A. R., Andrushchenko, I., Neilson, J. D. & Whelan, C. Indices of stock status from the Canadian bluefin tuna fishery. Collective Volume of Scientific Papers ICCAT 69 (2013).

35. 35.

Campana, S. E. et al. Recovery potential assessment for porbeagle (Lamna nasus) in Atlantic Canada. DFO Canadian Science Advisory Secretariat 2015/041 (2015).

36. 36.

DFO. Recovery potential assessment report on white sharks in Atlantic Canada. DFO Canadian Science Advisory Secretariat 2006/052 (2006).

37. 37.

COSEWIC. Assessment and status report on the shortfin mako Isurus oxyrinchus, Atlantic population, in Canada. Committee on the Status of Endangered Wildlife in Canada (2017).

38. 38.

Campana, S. E., Joyce, W. & Fowler, M. Subtropical pupping ground for a cold-water shark. Can. J. Fish. Aquat. Sci. 67, 769–773 (2010).

39. 39.

Francis, M. P., Holdsworth, J. C. & Block, B. A. Life in the open ocean: seasonal migration and diel diving behaviour of Southern Hemisphere porbeagle sharks (Lamna nasus). Mar. Biol. 162, 2305–2323 (2015).

40. 40.

Biais, G. et al. Return migration patterns of porbeagle shark (Lamna nasus) in the Northeast Atlantic: implications for stock range and structure. ICES J. Mar. Sci. 74, 1268–1276 (2017).

41. 41.

Hammill, M. O., den Heyer, C. E. & Bowen, W. D. Grey seal population trends in Canadian Waters, 19602014. DFO Canadian Science Advisory Secretariat 2014/037 (2014).

42. 42.

Ó Cadhla, O. et al. Grey seal breeding population assessment in the Republic of Ireland, 2005 (2007).

43. 43.

Marshall, K. N., Stier, A. C., Samhouri, J. F., Kelly, R. P. & Ward, E. J. Conservation challenges of predator recovery. Conserv. Lett. 9, 70–78 (2016).

44. 44.

Hedger, R. D., Rikardsen, A. H. & Thorstad, E. B. Pop-up satellite archival tag effects on the diving behaviour, growth and survival of adult Atlantic salmon Salmo salar at sea. J. Fish Biol. 90, 294–310 (2017).

45. 45.

Crossin, G. T. et al. Acoustic telemetry and fisheries management. Ecol. Appl. 27, 1031–1049 (2017).

46. 46.

Jepsen, N., Thorstad, E. B., Havn, T. & Lucas, M. C. The use of external electronic tags on fish: an evaluation of tag retention and tagging effects. Anim. Biotelemetry 3, 49, https://doi.org/10.1186/s40317-015-0086-z (2015).

47. 47.

Carey, F. G., Kanwisher, J. W. & Stevens, E. D. Bluefin tuna warm their viscera during digestion. J. Exp. Biol. 109, 1–20 (1984).

48. 48.

Goldman, K. J., Anderson, S. D., Latour, R. J. & Musick, J. A. Homeothermy in adult salmon sharks, Lamna ditropis. Environ. Biol. Fishes 71, 403–411 (2004).

49. 49.

Ricker, W. Computation and interpretation of biological statistics of fish populations. Bulletin of the Fisheries Research Board of Canada (1975).

50. 50.

Hoolihan, J. P. et al. Evaluating post-release behaviour modification in large pelagic fish deployed with pop-up satellite archival tags. ICES J. Mar. Sci. 68, 880–889 (2011).

51. 51.

Humphries, N. E. et al. Environmental context explains Lévy and Brownian movement patterns of marine predators. Nature 465, 1066–1069 (2010).

## Acknowledgements

We thank Alta Laksefiskeri Interessentskap, Atlantic Salmon Conservation Foundation, Danish Rod and Net License Fund, Inland Fisheries Ireland, Miramichi Salmon Association, Research Council of Norway project 280308 SeaSalar, Tromsø Research Foundation, and Xunta de Galicia for supporting and funding this project. We also thank all staff, students, and volunteers who have contributed to the fieldwork, and Timothy F. Sheehan and Jaakko Erkinaro for insightful discussion on an earlier version of the manuscript.

## Author information

Authors

### Contributions

A.H.R., J.C., K.A., P.G., P.C.J. and E.B.T. designed the study; S.E.C. and M.J.S.W. provided additional data; J.F.S. conceived the idea for the manuscript; J.F.S. analyzed the data with assistance from S.E.C. J.F.S., S.E.C., A.H.R. and D.R. interpreted the results; J.F.S. wrote the manuscript with input from E.B.T., S.E.C., D.R., A.H.R., J.C., K.A., P.G. and M.J.W.S.

### Corresponding author

Correspondence to John Fredrik Strøm.

## Ethics declarations

### Competing Interests

The authors declare no competing interests.

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

## Rights and permissions

Reprints and Permissions

Strøm, J.F., Rikardsen, A.H., Campana, S.E. et al. Ocean predation and mortality of adult Atlantic salmon. Sci Rep 9, 7890 (2019). https://doi.org/10.1038/s41598-019-44041-5

• Accepted:

• Published:

• ### Distinguishing discard mortality from natural mortality in field experiments based on electronic tagging

• Hugues P. Benoît
• , Jeff Kneebone
• , Sean R. Tracey
• , Diego Bernal
• , Klaas Hartmann
•  & Walt Golet

Fisheries Research (2020)

• ### Behavior and thermal environment of Chinook salmon Oncorhynchus tshawytscha in the North Pacific Ocean, elucidated from pop-up satellite archival tags

• Michael B. Courtney
• , Mark D. Evans
• , John F. Strøm
• , Audun H. Rikardsen
•  & Andrew C. Seitz

Environmental Biology of Fishes (2019)

• ### Diet and feeding strategy of Northeast Atlantic mackerel (Scombrus scomber) in Icelandic waters

• Cecilia Kvaavik
• , Anna Kristín Daníelsdóttir
• , Gudrún Marteinsdóttir
•  & Geir Ottersen

PLOS ONE (2019)

• ### Overlooked aspects of the Salmo salar and Salmo trutta lifecycles

• Kim Birnie-Gauvin
•  & Kim Aarestrup

Reviews in Fish Biology and Fisheries (2019)

• ### Nutritional correlates of the overwintering and seaward migratory decisions and long-term survival of post-spawning Atlantic salmon

• X Bordeleau
• , B G Hatcher
• , S Denny
• , F G Whoriskey
• , D A Patterson
•  & G T Crossin

Conservation Physiology (2019)

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.