Differential use of salmon by vertebrate consumers: implications for conservation

Study area

Spawning ground monitoring was conducted in the Chilkat and Chilkoot drainages near Haines, Alaska (Fig. 1) from June to November 2011. Monitoring of individual carcasses took place between August and October in both 2012 and 2013. The Chilkoot River flows less than 1 km from Chilkoot Lake before reaching the ocean. Chilkoot Lake is a glacially turbid lake, approximately 6 km long and 2 km wide. Primarily sockeye (Oncorhynchus nerka), but also coho (Oncorhynchus kisutch), spawn on the shores of the lake, and pink salmon (Oncorhynchus gorbuscha) spawn in the lower river and lake. Sockeye and coho also spawn in the river upstream of the lake.

The Chilkat drainage is a larger river system with multiple tributaries. The river is braided through a wide valley and meets the Tsirku and Klehini rivers 21 miles from the ocean at an area known as the Council Grounds. Ground water wells up from an alluvial fan at the intersection of these rivers, which prevents the river from freezing. A late chum salmon (Oncorhynchus keta) run spawns in the Chilkat River near the Council Grounds from October to December and in the lower Klehini River earlier in September–October. Pink salmon spawn in creeks 10 miles and 18 miles from Haines and in parts of the upper Chilkat River from August to early September. Sockeye salmon spawn in pools in the upper Chilkat River, Mosquito Lake, and in Chilkat Lake with early and late runs spanning June through October. Coho spawn in lakes and streams throughout the watershed but not in dense aggregations. Coho salmon persist in small streams into January. We identified all run timings through direct observation while visiting salmon spawning areas identified by Alaska Department of Fish and Game. The availability of chum and coho salmon late in the year draws a large congregation of bald eagles to the Chilkat River. Black-billed magpies (Pica hudsonia), common ravens (Corvus corax), mew gulls (Larus canus), and glaucous-winged gulls (Larus glaucescens) are also abundant and feed alongside the eagles.

Spawning ground monitoring

We used twenty motion-activated infrared cameras (Bushnell Trophycam) to monitor wildlife activity on pink, chum, and sockeye spawning grounds from August to November 2011 (Figs. 1 and 2). Motion-activated cameras on spawning areas may produce biased results because larger-bodied animals can trigger the camera at a greater distance in the viewshed. We elected not to sample on a systematic interval because such sampling is likely to miss infrequent visitors unless the sampling frequency is very high (i.e., nearly continuous monitoring), which would produce an intractably large number of photos for a study of this scale. We were not concerned about any bias introduced by motion-activated cameras because this research was designed to understand qualitative and broad-scale natural history patterns.

We classified stream morphology into creeks, pools, flats, and lakes. Creeks are small streams with forest cover, which includes the pink salmon spawning grounds at 10 mi and 18 mi Creek, and chum salmon spawning grounds at Herman Creek. Pools are slow moving shallow offshoots of the Chilkat River (i.e., alcoves), including Bear Flats and Mule Meadows. Flats are expansive seasonally-flooded areas of braided river with no forest cover on the Chilkat and Klehini Rivers. The lake category included the lakeshore spawners of Chilkoot Lake and Mosquito Lake (Fig. 1). We monitored dirt roads and trails on the Kelsall road system, which parallels the upper Chilkat River and branches to follow the Kelsall River and Nataga Creek tributaries (Fig. 1). We placed camera traps on roads and trails in early summer to provide an index of relative abundance of terrestrial carnivores such as black bears, brown bears, coyotes, lynx, and wolves. This method may introduce biases if some carnivores select for or against roads and trails, but encounter rates serve as a useful index of abundance. All cameras were set to take three pictures when triggered with a three second delay between successive triggers. To avoid overestimating visitation rates when the same individual or group foraged in front of the camera for an extended period, we post-processed the camera data to identify unique visitations. We defined unique group visitations (where “unique” refers to a unique encounter rather than a unique individual) as visits with a greater than two minute delay between the last photograph from one visit to the first photograph of the next. We weighted each unique group visit by the observed group size to estimate the number of individual encounters (unique visits weighted by group size per camera-day). We chose a short two-minute delay because we often observed one group of animals replace another in quick succession (e.g., subdominant sow with cubs replaced by another sow).

Salmon carcass monitoring

We used motion-activated infrared cameras (Bushnell Trophycam) to monitor wildlife visitations to, and feeding activity on, individual salmon carcasses at two mid-season sockeye runs (Chilkoot Lake and Mule Meadows, August and September) and one late season chum salmon run (Herman Creek, September and October) in both 2012 and 2013. At each site we erected a grid of eight cameras, four cameras 15 m from spawning grounds and four cameras 50 m from spawning grounds, with 150 m lateral spacing between cameras. Each camera was baited with a single salmon carcass staked to the ground. Salmon carcasses were collected from the adjacent spawning grounds and were monitoring during the period of the salmon run. All cameras were set to take three pictures when triggered with a one-minute delay between successive triggers. Cameras were checked weekly and carcasses were replaced if missing or decomposed. We defined unique visitations as visits with a greater than five minute delay between the last photograph from one visit and the first photograph of the next on any one camera at the site.

Spawning ground monitoring, 2011

Brown bears foraged extensively on all salmon runs regardless of species, run timing, and stream morphology (Figs. 2–4), including lakeshore spawners. Ravens, which are resident forest birds, were the only birds to feed extensively on early runs of pink salmon, which were also consumed by mustelids (mink and marten) and coyotes (Fig. 4A), but ravens were not observed at 10 mile creek despite an abundance of carrion (Fig. 3). Ravens also generally fed more than other avian scavengers on sockeye spawning grounds (Fig. 4A), but they were outnumbered at Chilkoot Lake, where eagles were more often observed, and Bear Flats, where magpies were observed slightly more frequently (Fig. 3). We speculate that ravens selected against the relatively coastal spawning areas at Chilkoot Lake and 10 mile creek in favor of concurrent inland spawning areas.

Migratory avian scavengers, including eagles, gulls, and magpies did not feed on early pink salmon runs (Figs. 3 and 4A). This was likely due to a combination of early run timing and avoidance of small streams when salmon are available elsewhere at sites with more suitable stream morphology. A concurrent larger pink salmon run occurs on the more expansive Chilkoot River where large numbers of gulls and eagles feed on pink salmon, but we did not monitor this run with cameras because the river is heavily used by humans. The absence of birds, other than ravens, at early pink salmon runs was followed by a high concentration of eagle and gull activity at late chum salmon runs (Figs. 3 and 4). However, avian scavengers fed on salmon less often than brown bears at the late chum salmon run at Herman Creek. We speculate that stream morphology and run timing both influenced where these consumers preferentially fed on salmon (Figs. 3 and 4). This was particularly evident by the very few gulls observed at Herman Creek relative to Klehini Flats and Chilkat Flats.

Brown bears were observed disproportionately more on salmon spawning grounds than expected by their abundance. Black bears were the most commonly observed animal on roads and trails during summer 2011, followed by coyotes, brown bears, wolves and lynx (Fig. 5). Black bears, which are dominant salmon consumers in many systems without brown bears, were notably absent on salmon spawning grounds (Fortin et al., 2007) (Fig. 5), which suggests that risk associated with interference competition by brown bears is strong enough to prevent black bear consumption of this energetically profitable resource. Similarly, coyotes were much more abundant than wolves, but wolves were observed as frequently as coyotes on salmon spawning grounds. Lynx were not observed on spawning grounds (Fig. 5).

Salmon carcass monitoring, 2012–2013

A wide variety of species were observed both visiting and consuming salmon carcasses in both 2012 and 2013 (Fig. 6). Encounter rates of species varied among sites and between years (Fig. 7). We observed no differences between visitations at carcasses near to (15 m) and farther from (50 m) spawning grounds for any species (all p > 0.2). As with spawning ground monitoring in 2011, brown bears were the dominant visitors to all carcasses regardless of run timing. Although avian activity overall was observed to be higher at late runs at Herman Creek than mid-season runs at Chilkoot Lake and Mule Meadows, avian encounters were much lower on individual salmon carcasses in 2012 and 2013 than on spawning grounds in 2011. As salmon carcass sites were either 15 or 50 m from spawning grounds and often under forest cover, this could suggest that avian scavenging on partially-consumed salmon carcasses deposited by bears might be limited to larger, open areas immediately adjacent to spawning grounds or in sparsely or unforested habitats like river flats.

Visitation of wildlife to salmon carcasses occurred consistently at each site, but consumption of carcasses was observed less frequently. While small carnivores such as mink and marten and avian scavengers were observed feeding on individual carcasses at all sites throughout the duration of the carcass monitoring, bears rarely fed on carcasses until the end of the late chum salmon run at Herman Creek. Since salmon carcasses were used to bait camera stations rather than live salmon, this could indicate a preference for predation over scavenging, or for consumption of fresher fish until the point at which salmon becomes limiting. The observation that bears frequently return to salmon carcasses after initially high grading on the most calorie-rich portion of the fish suggests that estimates of salmon consumption by bears are likely biased low (Gende, Quinn & Willson, 2001).

The extensive use of all salmon resources by brown bears suggests that bears are benefiting from a diversity of run timing by moving between asynchronous spawning aggregations to maximize their nutritional intake (Schindler et al., 2013). It is possible that relaxing harvests on lower value salmon species such as pink salmon (2014 Alaska ex-vessel value of $0.28/lb) and chum ($0.64/lb), while fishing higher value species such as king ($4.27/lb), and sockeye ($1.75/lb) for maximum sustainable yield would be an effective strategy to increase bear population productivity with less economic impact. This seems particularly plausible because pink and chum spawn early and late in the season, respectively, which extends the temporal availability of salmon biomass. However, these species typically do not permeate as far into watersheds as other salmon species, restricting access to more interior salmon predators and scavengers. It is also important to note that migration timing, when salmon are available for commercial harvest, and timing of spawning, when salmon are available to terrestrial consumers, do not always closely correspond (Boatright, Quinn & Hilborn, 2004; Doctor et al., 2010), so a focus on early or midseason runs for commercial harvest might not result in increased escapement of late spawners.

The complete absence of eagles and other avian scavengers on early pink salmon runs suggests that regional salmon availability far exceeds their energy requirements during this time of year. These avian scavengers congregate in enormous numbers at late chum salmon runs when salmon become a limiting resource (Elliott et al., 2011; ∼3,500 observed at our field site by aerial counts). The Chilkat River eagle gathering from October to January is the earliest of these congregations and the farthest north. The Squamish and Harrison river systems in southern British Columbia also support large eagle congregations that peak in January. Other much smaller congregations continue in south-coastal British Columbia and Northern Washington State though April. These few late chum salmon runs are disproportionately important resources for the larger panmictic eagle population in the Pacific Northwest, Alaska, and Western Canada (Elliott et al., 2011).

An obvious question is why other terrestrial vertebrates use salmon so much less than brown bears and avian scavengers. Bears maximize their annual energy balance by depositing fat during salmon abundance and hibernating during food scarcity, and migratory avian scavengers can track asynchronous pulses of salmon on vast spatial scales. We speculate that an inability to employ these life history strategies prevents other salmon consumers from reaching the atypical population densities (i.e., relative to elsewhere in their range) reached by bears and mobile avian scavengers in the presence of salmon. Winter food availability rather than salmon may limit the populations of most terrestrial carnivores, but the importance of salmon to each species is likely to be idiosyncratic. Wolves, for example, are more successful at hunting ungulate prey in winter than during summer, such that salmon can lead to apparent competition by maintaining wolves at high density even when ungulate biomass is low (Adams et al., 2010). Black bears do possess the hibernation strategy to effectively exploit pulsed salmon resources, and do so in many river systems without brown bears, but interference competition with brown bears prevents salmon consumption at our field site. It is unclear to what degree interspecific competition affects salmon consumption by other species, but is plausible that coyotes, which are locally abundant (Fig. 4), were observed infrequently due to interference competition from wolves. For small-bodied scavengers, such as mink and marten, the spatial distribution of salmon, rather than biomass per se, may be a more important driver of population productivity because even relatively small quantities of salmon may support the low population density of resident individuals. However, this depends on the accessibility of salmon carcasses, which can be quickly flushed from systems in the absence of large woody debris (Cederholm et al., 1999), or deposition of partially consumed carcasses by bears (Helfield & Naiman, 2006). Additionally, small carnivores may focus more on salmon carcasses when small mammals, their alternative prey, are rare, because rodents have much higher energy densities than do spawned out salmon carcasses (∼7–8 kJ/g for rodents compared to ∼3 kJ/g for spawned out salmon Cox & Secor, 2007; Hendry & Berg, 1999).

Understanding how salmon use is influenced by consumer life history, run timing and stream morphology is an important first step toward integrating wildlife needs into salmon management as part of an ongoing paradigmatic shift toward ecosystem-based fisheries management (Levin et al., 2009). Although there are quantitative differences in the nutritional quality of salmon and their accessibility, all salmon on the landscape were heavily used by bears. In contrast, the activity pattern of bald eagles suggests that fisheries management that considers their nutritional requirements might instead focus on increasing escapements at late chum salmon runs where eagles congregate in Southeast Alaska, British Columbia, and Northern Washington. Similarly, resource extraction that threatens late chum salmon runs would likely have a disproportionately large impact on regional eagle populations.