Seals (Phocidae) are known predators of Atlantic Salmon Salmo salar from their entry into the estuary as smolts until their return to freshwater as adults. We developed a written protocol for identifying seal-induced injuries on adult Atlantic Salmon returning to Maine rivers. The protocol, which includes photos and category definitions, has been used since 2006 by Maine Department of Marine Resources biologists when handling adult Atlantic Salmon at the Veazie Dam fishway on the Penobscot River and at other adult capture facilities in Maine. The written protocol has ensured that reporting is consistent among years and rivers; photos of wounds serve as quality assurance for the protocol. For adults returning to the Penobscot River in 2006 and 2007, seal-induced injuries were more likely to be found on two-sea-winter fish; to appear as gashes or arched wounds; to be open rather than healed; and to occur on the center third of the body below the lateral line. Larger two-sea-winter fish (>67 cm FL) returning to the Penobscot River early in the run (May-early July) were more likely to have a seal-induced injury. Injury rates increased from 1978 to 2010; this increase was correlated with seal aerial counts for the Gulf of Maine and Penobscot Bay. From 2006 to 2010, the annual seal-induced injury rate for Atlantic Salmon in six Maine rivers (including the Penobscot River) ranged from 0.00 to 0.30 across rivers. Rates of injury on Atlantic Salmon for all years combined did not differ among the six rivers, but low power likely affected our ability to detect any differences.
Atlantic Salmon Salmo salar face predators throughout the marine phase of their life cycle; they may be consumed from the time they enter estuaries as smolts (Greenstreet et al. 1993) until their return to freshwater as adults (Carter et al. 2001). Upon entry into the marine environment, the suite of predators that may be encountered by Atlantic Salmon includes other fishes, cetaceans, birds, and seals (Phocidae; Cairns 2002). The ability to survive predation attempts in the ocean is paramount to the persistence of Atlantic Salmon populations (Anthony 1994).
Seals are common throughout the North Atlantic (Mansfield 1967), where they prey on many species, including Atlantic Salmon (Pierce et al. 1989; Hammill and Stenson 2000; Carter et al. 2001; Bowen et al. 2002; Cairns 2003). Although Atlantic Salmon may represent a small proportion of an individual seal's diet (Pierce et al. 1989; Hammill and Stenson 2000; Carter et al. 2001; Bowen et al. 2002), seals in the North Atlantic consume tens of thousands of tons of fish annually (Hammill and Stenson 2000). Thus, even if the Atlantic Salmon is not an important prey item for seals, the demographic effect of seal predation on Atlantic Salmon may be substantial (Cairns 2001).
Seal abundance may increase locally when Atlantic Salmon return to a river (Middlemas et al. 2006). However, overall seal population increases are more likely related to abundances of other prey types, with seals opportunistically preying on salmon. Thus, at low Atlantic Salmon numbers, predation from locally abundant seals may further reduce populations of returning adult salmon (Middlemas et al. 2003), particularly in the absence of abundant alternative prey (e.g., anadromous alosines) in estuaries (Saunders et al. 2006).
The Gulf of Maine (GOM) distinct population segment (DPS) of Atlantic Salmon consists of all anadromous Atlantic Salmon populations from the Androscoggin River to the Dennys River (USOFR 2009) in Maine. The number of Atlantic Salmon returning to rivers in the GOM DPS decreased from the mid-1980s to the early 2000s (Fay et al. 2006; USASAC 2012). Concurrently, coastal seal populations have increased in the Northwest Atlantic (Waring et al. 2007) and in the GOM (Gilbert et al. 2005). Richardson (1976) estimated the population of harbor seals Phoca vitulina in Maine to be approximately 5,000 animals in 1973 and 1974, just after passage of the Marine Mammal Protection Act in 1972. The harbor seal population along the coast of Maine increased 6.6% annually from 1981 to 2001, reaching 99,340 animals in 2001 (Gilbert et al. 2005). In the same period, the observed number of grey seals Halichoerus grypus increased from 0 to 1,731 animals (Gilbert et al. 2005).
In Maine, the fisheries agency that is responsible for Atlantic Salmon management has been documenting the number of seal-induced injuries observed among adult Atlantic Salmon returning to the Penobscot River since 1978 (MASC 2011). Before 2006, annual training on the identification of seal-induced injury was conducted by experienced staff using live, injured salmon; however, there was no pictorial or written protocol. A written, photo-documented protocol based on the methods of Harmon et al. (1994) and Thompson and Mackay (1999) was developed in 2006 and has since been in use at adult Atlantic Salmon capture facilities operated by the Maine Department of Marine Resources (MDMR). In this study, we used information collected by MDMR on the Penobscot River to (1) describe types, locations, and timing (recent or healed) of observed injuries on returning adult Atlantic Salmon; (2) describe which characteristics of returning adults were associated with the presence of seal-induced injuries; (3) document the proportion of returning adults with one or more seal-induced injuries; and (4) demonstrate the trend in seal-induced injury rates among adult Atlantic Salmon from 1978 to 2010 and the relationship between the injury rates and harbor seal aerial counts. Finally, we compared the rates of seal-injured Atlantic Salmon among six Maine rivers (including the Penobscot River) from 2006 to 2010, when the injury identification protocol was in use.
METHODS
The Penobscot River enters the GOM in mid-coast Maine (Figure 1) and contains a population of Atlantic Salmon that was listed as endangered after publication of the 2006 Status Review for Atlantic Salmon (Fay et al. 2006; USOFR 2006, 2009). The Penobscot River has the greatest number of returning Atlantic Salmon adults in the United States (Fay et al. 2006). The MDMR operates a trap at Veazie Dam on the Penobscot River (river kilometer 47; Figure 1) to count returning adults. Adult Atlantic Salmon returning to the river from 2006 to 2010 were captured at the trap on the Veazie Dam fishway by MDMR personnel and were screened for injuries and other abnormalities during daily trap operation. Recorded injuries included avian predator strikes, hooking wounds from incidental or illegal exploitation, lacerations, lamprey wounds, and seal bites. At first capture, FL (cm) was measured; observed fin clips, visual implant elastomer marks, and tags were recorded; a scale sample was removed; fin erosion was scored; and sex was assigned visually. Atlantic Salmon were assigned an origin (hatchery or natural) based on marks or fin erosion scores and freshwater growth patterns on scales (if not marked), sea age was determined from scale pattern analysis, and sex was assigned using secondary characteristics or was known (i.e., for fish spawned at Craig Brook National Fish Hatchery). In addition, during 2006 and 2007, a separate field form was used if injuries were likely to have been caused by a seal; this form was used to record (1) whether the injury was healed or open, (2) wound types within each injury (Table 1), and (3) the injury's location on the body. To verify the data recorded, photographs of the injuries were taken. An individual could have one or more injuries, healed and/or open injuries, and one or multiple wound types within an injury. All field forms and photos were reviewed to ensure that injury age and wound type(s) were appropriately assigned.
FIGURE 1.
Map of Maine rivers (1 = Dennys; 2 = Narraguagus; 3 = Penobscot; 4 = Kennebec; 5 = Androscoggin) with Atlantic Salmon belonging to the Gulf of Maine (GOM) distinct population segment (DPS). The Saco River (river 6) is not within the GOM DPS but was included in the analysis of seal-induced injury rates. The downstream-most trap used to capture adult Atlantic Salmon on each river is represented by an open circle.
TABLE 1.
Seal injury types and associated descriptions used to identify sealinduced injuries on Atlantic Salmon returning to the Penobscot River, Maine, in 2006 and 2007. For the purposes of analysis, type III injuries were considered seal bites, as photos indicated that these injuries were likely variations of type I and type II injuries.
The written protocol for identifying injuries was also used for fish captured on the Dennys, Narraguagus, Kennebec, Androscoggin, and Saco rivers from 2006 to 2010 (MASC 2011; Figure 1), but separate field forms or photos were not produced. On the Penobscot River after 2007, photos of injuries were taken to augment the examples included in the protocol, but separate seal injury field forms were not completed. Historical information on seal-induced injuries (1978–2005) was used for long-term analyses.
Logistic regression was used to evaluate the influence of year, day of year, sea age, sex, origin, and FL on the likelihood that an Atlantic Salmon returning to the Penobscot River from 2006 to 2010 would have a seal-induced injury. Separate models for each sea age were used to evaluate the influence of FL because length in Atlantic Salmon is not independent of sea age. One-sea-winter (1SW) and two-sea-winter (2SW) Atlantic Salmon overlapped in FL over the range of 58–64 cm. The first main-effects model for seal-induced injury was
where sea age = number of winters at sea (1SW versus ≥2SW); day of year = continuous variable (days since January 1 of the year); origin = rearing history (hatchery versus natural); sex = gender (female versus male); and year = a continuous variable (study year, from 2006 to 2010). The model for 2SW fish included FL and excluded sea age, while the model for 1SW fish included FL and excluded sex (all were male) and sea age. Models were based on the fewest independent variables and interactions that produced a nonsignificant likelihood ratio in the ANOVA table.
Linear regression was used to demonstrate the trend in sealinduced injury rates (weighted by run size) from 1978 to 2010, and Spearman’s rank correlation was used to describe the relationship between seal-induced injury rates and harbor seal aerial counts. Nonparametric analysis (i.e., Kruskal—Wallis test) was used to test for differences in Atlantic Salmon injury rates among rivers. All statistics were calculated with R software (R Core Development Team 2014); an α of 0.05 was used to assess significance for all analyses.
RESULTS
From 2006 to 2010, the number of adult Atlantic Salmon returning to the Penobscot River ranged from 916 to 2,115, and the proportion of returning adults with seal-induced injuries ranged from 0.016 to 0.042 (Table 2). Most of the returning adults each year had spent at least 2 years at sea and were of hatchery origin. One-sea-winter individuals were often a large component of the run and accounted for a greater number of males than females being captured each year, since 1SW Atlantic Salmon are almost always male.
In 2006, wound type and associated data were recorded for all 27 Atlantic Salmon with seal-induced injuries, and photos were taken of the injuries. In 2007, injuries were again identified based on the protocol, but photos and wound type data were available for only 16 of the 23 seal-injured fish. Thus, detailed data were available for a total of 47 injuries (Tables 3, 4). Most of the injured fish had only one injury; however, there were four 2SW hatchery fish that each had two seal-induced injuries (one healed and one open). Less than half of the injuries photographed were healed at the time of sampling (Table 3). No seal-induced injuries were seen on the heads of Atlantic Salmon. Thirty-nine of the injuries were situated below the lateral line, and 32 of these were in the middle portion of the body (Table 3). Within 17 of the injuries, several wound types were identified (Table 4). Most of the injuries either were wound type IV or included wound type IV (n = 21); wound type I (n = 16) was the second most prevalent type.
TABLE 2.
Number of returning adult Atlantic Salmon in the Penobscot River during 2006–2010 and the proportion of adults with seal-induced injuries (1SW = one-sea-winter fish; 2SW = two-sea-winter and older fish; HO = hatchery-origin fish; NO = natural-origin fish; M = males; F = females).
All three logistic models used to describe the likelihood of an Atlantic Salmon having a seal-induced injury had a nonsignificant residual (df = 7,341, 5,360, 1,964; P = 1.00) and were significantly different from the null model (df = 5, 5, 4; P < 0.0001). In the first model (i.e., for all adults combined), sea age and day of year were the only significant variables (Table 5). Thus, seal injury rates were similar between males and females and between natural and hatchery-reared fish and exhibited no trend over the study period (Table 2). The model for 2SW Atlantic Salmon had two significant variables (day of year and FL) and three nonsignificant variables (origin, sex, and year; Table 5). Two-sea-winter fish were more likely to have a seal-induced injury at the beginning of the run (Figure 2a). This coincided with a high proportion of the entire run passing Veazie Dam and the observation of the greatest numbers of fish with seal-induced injuries (Figure 2b). The likelihood of being injured differed among 2SW individuals based on FL, with the likelihood increasing as FL increased (Figure 3). In the 1SW model, the intercept and year were significant variables, and day of year and FL were not significant. Injury rates for smaller fish were low and increased over the study period (Table 2).
TABLE 3.
Anatomical locations of seal-induced injuries on Atlantic Salmon returning to the Penobscot River in 2006 and 2007 (lateral line = the injury's position relative to the lateral line).
From 1978 to 2010, the annual seal-induced injury rates for Atlantic Salmon returning to the Penobscot River ranged from a low of 0.001 in 1981 to a high of 0.11 in 1995 (Figure 4). The seal-induced injury rate increased over time from 1978 to 2010 (linear regression; t = -3.217, P = 0.003; Figure 4). Sealinduced injury rates increased with increasing aerial counts for harbor seals along the coast of Maine (one-sided Spearman's rank correlation: coefficient rs = 0.89, P = 0.02) and in Penobscot Bay (rs = 0.90, P = 0.04).
FIGURE 2.
(a) Univariate logistic model (logit [yes/no] = B 0 + B 1 [day of year]) probability (solid line; with 95% confidence interval, dashed lines) of two-sea-winter (2SW) Atlantic Salmon having seal-induced injuries; and (b) the number of Atlantic Salmon with seal-induced injuries (black bars), and the proportion of all adults captured at the Veazie Dam fishway on the Penobscot River (solid-gray line) each day during 2006–2010.
TABLE 4.
Number and combinations of injury types observed on Atlantic Salmon with seal-induced injuries in the Penobscot River, 2006 and 2007. Refer to Table 1 for injury type definitions.
TABLE 5.
Summary of logistic regression models describing the presence or absence of seal-induced injuries on adult Atlantic Salmon returning to the Penobscot River, 2006–2010 (Z = Wald z-statistic; AIC = Akaike's information criterion).
FIGURE 3.
Univariate logistic model (logit [yes/no] = B 0 + B 1[FL]) probability (solid line; with 95% confidence interval, dashed lines) of two-sea-winter Atlantic Salmon having seal-induced injuries based on FL for adults returning to the Penobscot River, 2006–2010.
The annual seal-induced injury rate for Atlantic Salmon among the six Maine rivers ranged from 0.00 to 0.30 during 2006–2010 (Figure 5). For all years combined, there was no difference in injury rates among rivers when the injury identification protocol was used (Kruskal—Wallis test: H = 5.865, P = 0.32). However, the pattern in observed injury rates was markedly different among the rivers: in the Penobscot River, seal-induced injuries were observed on returning Atlantic Salmon during all years, whereas the other five rivers had no seal-induced injuries observed during 2 or more of the 5 years (Figure 5). The Dennys River in particular was notable in having 3 years with a relatively high proportion of returning adults exhibiting injuries (≥0.20) and 2 years with no injured adults observed.
FIGURE 4.
Proportion of returning adult Atlantic Salmon with seal-induced injuries in the Penobscot River (solid circles with connecting line), annual number of adult Atlantic Salmon returning to the Penobscot River (dashed line), harbor seal aerial counts (divided by 10) for the coast of Maine (open squares; Richardson 1976; Gilbert et al. 2005), and harbor seal aerial counts for Penobscot Bay (open diamonds; Gilbert et al. 2005) from 1978 to 2010. Seal-induced injury data within the gray rectangle were collected using the protocol described in this paper.
DISCUSSION
The MDMR has used various methods since 1978 to identify seal-induced injuries on Atlantic Salmon returning to the Penobscot River. Since 2006, use of the written protocol and collecting photographs and data on wound types within an injury helped to verify that an injury was properly identified as representing a predation attempt by a seal. Deep gashes into the flesh (type IV) and arches (type I) were the most commonly recorded wounds and were the easiest to identify, even when injuries included multiple wound types. The locations of seal injuries on the bodies of Atlantic Salmon provide information on unsuccessful seal attacks. About two-thirds of the injuries were below the lateral line and on the middle portion of the fish's body, indicating that when seals attack Atlantic Salmon from directly below, escape is possible. There were fewer individuals with injuries on the caudal peduncle and tail. A fish that is attacked from directly behind may be less likely to escape than a fish that is attacked from below. None of the examined Atlantic Salmon had seal-induced injuries on the head. Thus, we infer that attacks on the head are typically lethal, occur only rarely, or both.
FIGURE 5.
Proportion of returning adult Atlantic Salmon with seal-induced injuries in six Maine Rivers, 2006–2010 (Penobscot River = solid circles with connecting line; Dennys River = solid triangles; Narraguagus River = solid squares; Kennebec River = solid diamonds; Androscoggin River = open diamonds; Saco River = open triangles).
Atlantic Salmon migration involves relatively low numbers of fish passing through large expanses of the North Atlantic (Dadswell et al. 2010), with seal predation possible both in nearshore areas and in the open ocean (Hammill and Stenson 2000). We thought that information on whether the injury was open or healed could be used to infer where the attack might have occurred (i.e., in the river, estuary, or ocean). Epidermal migration closes wounds on Atlantic Salmon within 8 h (Roubal and Bullock 1988), with healing occurring over 1–3 months. Most (26) of the 47 injuries examined in 2006 and 2007 were classified as open, meaning that they were less than 1 h old to 3 months old. Because wound closure is temperature dependent (Anderson and Roberts 1975), open injuries apparent earlier in the season could have occurred farther from the Veazie Dam fishway than open injuries that were seen when river temperatures were higher. Davidsen et al. (2013) observed adult Atlantic Salmon in Norway traveling along the coastline when returning to their river of origin. Thus, it seems reasonable that many of the injuries observed in our study likely occurred during the spring or early summer as the Atlantic Salmon migrated through the GOM and along the coastline before entering the Penobscot River.
Similar to the results of Fryer (1998) and Naughton et al. (2011), who studied both sea lion-induced (Otariidae) and seal-induced injuries on Pacific Salmon Oncorhynchus spp., we observed that 2SW and older Atlantic Salmon exhibited most of the seal-induced injuries (Table 2). It is possible that (1) seals selectively attack larger Atlantic Salmon, (2) smaller fish escape attacks at a lower rate than larger fish, or (3) a combination of the two. We also observed strong seasonality in the returns of injured 2SW Atlantic Salmon, with earlier-arriving adults having a greater likelihood of bearing seal-induced injuries, whereas no injuries were documented after July 21. This pattern of higher risk earlier in the migration period is similar to that observed for Chinook Salmon O. tshawytscha in the Columbia River by Keefer et al. (2012); those authors found that later in the spring, the fish had a lower predation risk, likely due to intraspecific prey swamping. At the present abundance levels of Atlantic Salmon, intraspecific prey swamping via 2SW fish is unlikely. However, the potential for interspecific prey swamping (referred to as a “prey buffer effect” by Saunders et al. 2006) should be considered, particularly in light of recent evaluations by Grote et al. (2014), who provided compelling evidence that American Shad Alosa sapidissima inhabit the lower Penobscot River in greater numbers than previously known. For a prey swamping effect to occur, alternative prey resources must overlap in time and space. Spatially, there is overlap among Atlantic Salmon, American Shad, and Alewives A. pseudoharengus in the lower Penobscot River and estuary up to Veazie Dam (which was impassable to American Shad; the dam was removed in 2013). Temporal overlap is likely given that the median capture date of Atlantic Salmon at Veazie Dam in 2011 was June 15, and the median detection date of American Shad-sized targets in 2011 was June 14 (Grote et al. 2014). During 2009–2013, the median capture date for Alewives ranged from May 12 to May 28, with fish being captured at Veazie Dam from the beginning of May through the first week of June (MDMR, unpublished data); this period corresponds to the time when the number of Atlantic Salmon passing the dam begins to increase (Figure 2b). Annually from 2006 to 2010, Atlantic Salmon became increasingly rare potential prey items for seals in the lower river and estuary as they ascended the fishway and thus were no longer susceptible to seal predation. American Shad, however, could not pass Veazie Dam and therefore remained potential prey for seals throughout their spawning migration and during their postspawn downstream migration, which began in mid- to late June. For Atlantic Salmon, the risk of attack by seals declines through time within each year based on (1) the negative slope for the day-of-year variable in our regression models and (2) the observation of only one seal-induced injury after July 21 (Figure 2b). Future evaluations of prey selectivity, energy content per prey item, and seal abundance and behavior within the Penobscot River estuary could further elucidate these potential relationships.
Although seal-induced injury rates on adult Atlantic Salmon returning to the Penobscot River have typically been low (<0.04), they have increased since 1978. The numbers of harbor seals on the coast of Maine and in Penobscot Bay have also increased. Although we were able to document the number of Atlantic Salmon with seal-induced injuries over time, we do not know whether a higher proportion of returning adults with these injuries means that a higher or lower proportion of returns were taken by seals. This would require knowledge of seal success rates in capturing adult Atlantic Salmon of different lengths and at different Atlantic Salmon population sizes. With knowledge of capture success, injury rates might be used to calculate the numbers of Atlantic Salmon that are captured and eaten by seals in river estuaries. However, even if one assumed a within-season type III functional response for harbor seal predation on Atlantic Salmon (Middlemas et al. 2006), modeling the impact of seals on returning adult Atlantic Salmon in the Penobscot, Dennys, and Narraguagus River estuaries based on seal energetics would require data on seal abundance in each estuary (Butler et al. 2006) and the proportions of Atlantic Salmon in the seals' diets (Hammill and Stenson 2000).
It is not surprising that there was no significant difference in seal-induced injury rates among the six Maine rivers given the small sample size and resulting low statistical power. The Penobscot River had relatively stable and low injury rates from 2006 to 2010, whereas the Dennys, Narraguagus, and Saco rivers all exhibited more variable rates that ranged from zero in multiple years to greater than 0.09. High injury rates could be indicative of substantial predation and are of particular concern for rivers with severely depressed Atlantic Salmon populations. There are many variables that could explain the high injury rates on Atlantic Salmon in these rivers, including seal movement to the Maine coast from Massachusetts in the spring (Waring et al. 2006) and proximity of the river estuaries to seal haul-out locations and aquaculture facilities (Nelson et al. 2006); these relationships warrant further study.
The standardized protocol for identifying seal-induced injuries, which was developed in 2006 and is currently in use at Atlantic Salmon capture facilities in Maine, has ensured that reporting is consistent among years and rivers. Annual photos of common wound types within injuries are recommended as protocol quality assurance. Our analysis provides baseline injury rates for six Maine rivers and conveys insight into the types of seal-induced injuries that Atlantic Salmon may survive, the locations of these injuries, and variables that affect the likelihood of an Atlantic Salmon having a seal-induced injury. Future research that is focused on the interactions among seals, Atlantic Salmon, and other diadromous fishes; locating areas where seal predation is greatest; and describing seal attack rates and the rates at which Atlantic Salmon survive those attacks will be essential to advancing the knowledge of seal-salmon predator-prey dynamics.
ACKNOWLEDGMENTS
We are grateful to R. Dill, K. Dunham, and M. Simpson for their work coordinating data collection in 2006 and 2007, and to T. Trinko Lake for providing Figure 1. We also thank M. Keefer and an anonymous reviewer whose thoughtful comments helped improve the final manuscript. The data used in our analysis were collected by MDMR with funding from the National Marine Fisheries Service.
