Investigation of the 2018 thick-billed murre (Uria lomvia) die-off on St. Lawrence Island rules out food shortage as the cause.

Abstract

Die-offs of seabirds in Alaska have occurred with increased frequency since 2015. In 2018, on St. Lawrence Island, seabirds were reported washing up dead on beaches starting in late May, peaking in June, and continuing until early August. The cause of death was documented to be starvation, leading to the conclusion that a severe food shortage was to blame. We use physiology and colony-based observations to examine whether food shortage is a sufficient explanation for the die-off, or if evidence indicates an alternative cause of starvation such as disease. Specifically, we address what species were most affected, the timing of possible food shortages, and food shortage severity in a historical context. We found that thick-billed murres (Uria lomvia) were most affected by the die-off, making up 61% of all bird carcasses encountered during beach surveys. Thick-billed murre carcasses were proportionately more numerous (26:1) than would be expected based on ratios of thick-billed murres to co-occurring common murres (U. aalge) observed on breeding study plots (7:1). Concentrations of the stress hormone corticosterone, a reliable physiological indicator of nutritional stress, in thick-billed murre feathers grown in the fall indicate that foraging conditions in the northern Bering Sea were poor in the fall of 2017 and comparable in severity to those experienced by murres during the 1976–1977 Bering Sea regime shift. Concentrations of corticosterone in feathers grown during the pre-breeding molt indicate that foraging conditions in late winter 2018 were similar to previous years. The 2018 murre egg harvest in the village of Savoonga (on St. Lawrence Is.) was one-fifth the 1993–2012 average, and residents observed that fewer birds laid eggs in 2018. Exposure of thick-billed murres to nutritional stress in August, however, was no different in 2018 compared to 2016, 2017, and 2019, and was comparable to levels observed on St. George Island in 2003–2017. Prey abundance, measured by the National Oceanic and Atmospheric Administration in bottom-trawl surveys, was also similar in 2018 to 2017 and 2019, supporting the evidence that food was not scarce in the summer of 2018 in the vicinity of St. Lawrence Island. Of two moribund thick-billed murres collected at the end of the mortality event, one tested positive for a novel re-assortment H10 strain of avian influenza with Eurasian components, likely contracted during the non-breeding season. It is not currently known how widely spread infection of murres with the novel virus was, thus insufficient evidence exists to attribute the die-off to an outbreak of avian influenza. We conclude that food shortage alone is not an adequate explanation for the mortality of thick-billed murres in 2018, and highlight the importance of rapid response to mortality events in order to document alternative or confounding causes of mortality.

1. Introduction

In 2018 a mass mortality event of seabirds was observed in the northern Bering Sea, primarily on St. Lawrence Island (Duffy-Anderson et al., 2019; Romano et al., 2020). Seabird die-offs have become a regular event in Alaskan waters, with different species affected in different regions every year since 2015 (e.g. Jones et al., 2019; Piatt et al., 2020). In most of these events birds were severely emaciated, and above average sea surface temperatures have been implicated as the driving force behind the lack of food (Jones et al., 2019, 2018; Piatt et al., 2020). In these respects, the 2018 mortality event was similar; necropsied dead birds were emaciated (Romano et al., 2020), and the event followed a winter of record low sea ice (Duffy-Anderson et al., 2019). Other factors besides food shortages, however, including disease and toxins, may affect birds’ behavior, impact their ability to forage, and result in emaciation (e.g. Knowles et al., 2019).

The 2018 mortality event appeared to be unusual in its species composition: thick-billed murre (Uria lomvia) carcasses were most frequently encountered dead on beaches (Will et al. pers. obs.). There are two co-occurring murre species breeding on St. Lawrence Island colonies: thick-billed and common (U. aalge) murres. These co-occurring species breed sympatrically on cliff ledges, but partition prey resources. Common murres tend to eat pelagic forage fish, while thick-billed murres are generalist foragers, consuming benthic prey, large zooplankton, squid, and pelagic fish (Barger et al., 2016; Kokubun et al., 2018). In past mass mortality events, common murres were most often involved in food-related die-offs documented in both the Pacific and Atlantic Oceans (Piatt and Van Pelt, 1997; Furness and Tasker, 1999; Baily and Davenport, 2016; Piatt et al., 2020). Based on previous die-offs we would expect to find that common murres made up a larger proportion of dead birds on beaches. Alternatively, if both murre species were affected equally, we would expect to find that the ratio of dead thick-billed murres to common murres on beaches would reflect the ratio observed attending cliff faces prior to the die-off event.

The timing of the peak of the 2018 die-off was unusual compared to the timing of the peaks of other past mass mortality events. On St. Lawrence Island, large numbers of dead murres were reported washing up on beaches starting in late May and early June, at the beginning of the 2018 breeding season (Romano et al., 2020). Previously reported mortality events, both in Alaska and elsewhere in the Northern Hemisphere, have occurred primarily in mid-winter (Piatt and Van Pelt, 1997; Furness and Tasker, 1999; Baily and Davenport, 2016), or have included short-tailed shearwaters (Puffinus tenuirostris) which were overwintering from the Southern Hemisphere and occurred later in the summer (Baduini et al., 2001). One possible explanation for the unusual timing of the 2018 die-off is that birds arrived in the Arctic in poor condition after migrating from their lower latitude overwintering locations. If birds arrived in the Arctic in a weakened state, where they encountered low food availability, they may have been unable to expend the energy needed to capture prey. This would indicate that birds had also experienced food shortages on the wintering grounds, compounding poor conditions near their breeding colony. Non-lethal links between a bird’s experience in the late winter and their performance in the following breeding season have previously been observed in albatross (Thalassarche melanophris, T. chrysostoma) and Cassin’s auklet (Ptychoramphus aleuticus) (Sorensen et al., 2009; Crossin et al., 2017). We examined the species composition, timing, and evidence for food shortages to test whether food scarcity is an adequate explanation for the mass starvation of thick-billed murres in the northern Bering Sea (Romano et al., 2020), specifically in the area of St. Lawrence Island.

To assess whether and when birds were exposed to food shortages, we measured the concentration of the primary avian stress hormone corticosterone in blood and feather samples. These samples reflect three distinct time periods in an individual’s annual cycle: plasma collected at the time of the bird’s capture in the summer, and feathers grown when birds are away from the colony in the fall (primary feathers) but likely still in the northern Bering Sea and/or Chukchi Sea (Hatch et al., 2000; Kuletz et al., 2015, 2019, Takahashi et al., 2020) and in the late winter (throat feathers) when birds are likely outside of the northern Bering Sea, south of their breeding colonies (Takahashi et al., 2020). We have previously shown that corticosterone concentration in seabird plasma and feather samples is a reliable indicator of nutritional stress (Kitaysky et al., 2007, 2010 and references therein; reviewed by Sorenson et al., 2017; Will et al., 2018, 2019). Relatively high concentrations of corticosterone indicate exposure to food limitation, and can be used as an index of exposure to nutritional stress (i.e. individuals are not taking in adequate amounts of food to meet their energetic expenditures).

The 2018 die-off was originally reported by residents of St. Lawrence Island. Murres are important both as a food and cultural resource to the community of Savoonga, Alaska, located on the north coast of St. Lawrence Island in the northern Bering Sea (Fig 1). Between 1993 and 2018 adult murres made up, on average, 30% of the birds harvested by the community annually while murre eggs constituted, on average, 99.5% of the total eggs harvested (this study; Naves, 2015). Adult murres are harvested when they return to their colonies in the spring, and their eggs are harvested in late June during the first two weeks of egg laying. Due to the importance of seabirds as a subsistence resource, community members have a deep knowledge of what is normal for these birds, and expressed concern and alarm over the large numbers of birds washing up on their beaches in the early summer of 2018. Community members reported seeing murres near shore that they described as being sick and starving, displaying unusual swimming and diving behaviors. A previous die-off in November 2013 on the island was due to an outbreak of avian cholera that affected crested auklets (Aethia cristatella), thick-billed murres, northern fulmars (Fulmarus glacialis), common eiders (Somateria mollissima) and gulls (Bodenstein et al., 2015). Avian cholera has been attributed to a few other mass mortality events of seabirds and waterfowl at northern latitudes (Descamps et al., 2012; Wille et al., 2016). Warmer conditions in the Arctic may facilitate outbreaks of disease (Van Hemert et al., 2014), thus an alternative hypothesis to food shortage is that a disease outbreak occurred in early summer potentially compromising the foraging capabilities of murres, thus explaining both the unusual timing and why a single species dominated beach cast carcasses.

Figure 1.

Beach surveys on St. Lawrence Island were conducted between Kangi and Avighteq, the black triangles indicate these furthest western and eastern extents of our surveys. Colony-based work occurred at Kevipak, Myughi, Tanti and Tuqaghuk (solid circles). The 5 repeated beach transect locations are marked with inverted empty triangles. Study location (middle panel), includes points for the study area bottom-trawl survey stations used in the analysis of prey abundance. Occasionally weak and lethargic thick-billed murres were encountered during surveys (photo credit A. Will). Base map is adapted from Stamen Maps, ©OpenStreetMap contributors, openstreetmap.org/copyright, license CC BY-SA.

The 2018 die-off occurred in conjunction with oceanographic changes in the region, including low zooplankton availability (Nishizawa et al., 2020), the northward movement of southern Bering Sea fish species (Duffy-Anderson et al., 2019; Eisner et al., 2020), and abnormal current circulations (Basyuk and Zuenko, 2019, 2020). The 2018 die-off of seabirds may be a signal of a broader regime shift in the northern Bering Sea ecosystem (Grebmeier et al., 2006; Moore and Stabeno, 2015; Huntington et al., 2020). If birds died of starvation as a result of severe food shortage it may indicate a loss of a prey source, and highlight structural changes in the northern Bering Sea’s food web. In 1976–1977 the Bering Sea experienced a climate regime shift and a significant reorganization of the marine food web (Aydin and Mueter, 2007), followed by a steady decline in murre and kittiwake breeding populations on the Pribilof Islands (Byrd et al., 2008). Several studies indicate that birds experienced severe food shortages during the regime shift (Kitaysky et al., 2006; Renner et al., 2012; Will et al., 2018). To evaluate the severity of conditions in the northern Bering Sea in 2018 compared to historically severe conditions such as those in 1976–1977, we used museum specimens to compare exposure to nutritional stress in 2018 to what has occurred in the past. In addition, we examined the relative abundance of small forage size fish available to seabirds breeding on St. Lawrence Island colonies using standardized bottom trawl surveys during 2017–2019 (Will et al., 2020).

In this study we combine evidence from colony-based observations and physiology to examine which seabird species were most affected, whether poor foraging conditions occurred on the breeding, post-breeding, or overwintering grounds and the severity of possible food shortages in a historical context. We specifically compare the corticosterone concentrations in present day murres to those that experienced the 1970s Bering Sea regime shift to gain insight into the ecological significance of the 2018 seabird die-off. Finally, we report results from the few birds we were able to screen for disease to explore evidence for the alternative hypothesis that starvation was a product of disease rather than a food shortage.

2. Methods

2.1. Study Site

St. Lawrence Island, or Sivuqaq in Siberian Yupik, is located at the southern end of the Bering Strait (Fig 1), 58 km east of Russia’s Chukotka Peninsula and 264 km west of the Alaskan mainland. All field work was conducted from mid-July to late August in 2016–2019, and in late June 2019. In 2018 we arrived in time to observe the end of the mortality event (the die-off occurred late May thru early August, with a peak in June; Romano et al., 2020). Field sites are located at colonies near the village of Savoonga on the northern coast of the island: Kevipak (colony counts), Myughi, Tuqaghuk and Tanti (bird capture locations, Fig 1). The two Siberian Yupik villages (Savoonga and Gambell, Fig 1) on the island have an active subsistence lifestyle; 98% of the households in Savoonga harvest wild resources (Fall et al., 2013).

2.2. Colony Attendance

We conducted counts of common and thick-billed murres attending 11 study plots (established by L. Sheffield-Guy and A. Gall in 2003) once during the 2016 breeding season, twice during the 2017 breeding season and approximately every 7 days in 2018 and 2019. When two observers counted simultaneously, final counts had to match between both observers, when a single observer counted, counts had to match across three consecutive counts. To compare the ratio of live thick-billed murres to common murres attending the colony across years to the ratio of dead thick-billed murres to common murres found on beaches, we compared counts of thick-billed and common murres on study plots. These counts were conducted at approximately the same date and breeding stage across years: 12 Aug 2016, 31 Jul 2017, 25 Jul 2018, 28 Jul 2019.

2.3. Beach Surveys

From 19 July to 23 August 2018, we monitored bird carcasses on the nearest accessible shores from our monitoring colonies. In July 2018 we first surveyed 20 transects, including cobble, pebble, and sandy beaches along the shoreline from Kangi in the west to Avighteq to the east (Fig 1). Our initial surveys were an attempt to census the shoreline to gage the extent and severity of the die-off. We repeated five of the transects after 4 weeks to assess whether the die-off was still ongoing, resulting in a 2.625 km subsample of the shoreline (Fig 1). In 2019 we repeated surveys of these same five transects once each month in June, July, and August. The curvilinear distance walked on each shore portion was recorded. In 2018 our surveys totaled 16 km and in 2019 totaled 12.6 km of shoreline traveled. During surveys we walked slowly in a zig-zag pattern to cover the whole width of each beach. For each transect, we recorded the number of carcasses seen per species (or to the lowest possible level of taxonomic identification), and noted the GPS coordinates of start and end points. To avoid double counting carcasses, in 2018 we cleared the beaches of the carcasses we counted, and in 2019 we clipped the 10^th primary feather of each bird found.

2.4. Exposure to Nutritional Stress

We measured the concentration of corticosterone in thick-billed murres’ first primary feather, grown during the post-breeding molt in the fall, throat feathers, grown during the pre-breeding molt in late winter prior to when murres return to their breeding grounds (Gaston and Hipfner, 2000), and blood plasma collected from birds attending breeding ledges. Feathers were collected from live birds captured on the colony during late incubation and throughout the chick-rearing period, from dead birds found on beaches in 2018, and from museum specimens collected between St. Lawrence Island and Utqiagviq/Barrow, Alaska (see Suppl. Table 1 for catalog IDs, years, and sampling location). Museum samples were contributed by the University of Alaska Fairbanks’ Museum of the North, the Denver Museum of Natural History, the California Academy of Sciences, and the University of Washington’s Burke Museum. Additional samples were contributed by Alan Springer (University of Alaska, Fairbanks). Corticosterone has been shown to be stable in feathers over time (Bortolotti et al., 2009).

Corticosterone concentrations were measured using radioimmunoassay (Wingfield and Farner, 1975). Primary feathers were subsampled, 20–30 mm of the inner vane was removed from the rachis 20 mm from the feather’s tip and five whole throat feathers, with the calamus removed, were measured to the nearest mm. All feathers were weighed to the nearest 0.1 mg. Feathers were briefly washed in 500 ml isopropanol (High Performance Liquid Chromatography (HPLC) grade, Sigma- Aldrich; Will et al., 2018), then extracted in 3 ml of HPLC grade methanol (Fisher Scientific), and placed in a 50° C sonicating water bath for an hour, then left overnight in a 50° C water bath. Feathers were removed from the methanol post-extraction, recoveries were added, and samples were filtered via solid phase extraction with a methanol wash and air dried prior to adding phosphate-buffered saline with gelatin (PBSG) assay buffer (feather extraction procedure modified from Bortolotti et al., 2008). Blood samples were obtained from the alar vein within 3 minutes of bird capture, and kept cool until separated by centrifugation (< 12 hours post collection) and frozen at −20° C until analysis (Kitaysky et al., 1999). A volume of 20 μl of plasma was extracted using 4.5 ml of distilled di-chloromethane (Fisher Scientific), and air-dried prior to adding PBSG assay buffer. All samples were run in duplicate. Feather sample inter-assay coefficient of variation (CV): 4%, intra-assay CV: 0.94% and recoveries 94.7%. Plasma sample inter-assay CV: 1.94%, intra-assay CV: 1.16%, and recoveries 91.3% for 2016–2019 samples (St. George results reported previously by Kokubun et al., 2018).

To compare changes in body mass of breeding birds across years we weighed each bird we handled using a 2,000-gram spring-loaded Pesola scale. Thick-billed murres have high flight costs (Elliott et al., 2013), and do not generally accumulate large fat reserves (Jacobs et al., 2012), especially during the chick-rearing period when they actively reduce their mass (Croll et al., 1991). Body mass may be an adequate indication of condition for birds with low lipid stores (Jacobs et al., 2012; but see Schultner et al., 2013).

2.5. Prey Abundance

To assess possible changes in prey abundance that may have contributed to a food shortage we report results from Will et al., (2020) on fish abundance in the northern Bering Sea. Briefly, catch per unit effort (CPUE) data for fishes in the northern Bering Sea were obtained from the National Oceanic and Atmospheric Administration’s Alaska Fisheries Science Center Groundfish Assessment Program bottom trawl surveys (e.g. Lauth et al., 2019; Stevenson and Lauth, 2019) conducted during June-August of 2017, 2018, and 2019. Trawl data were sub-set by area, including only stations within a ~150 km radius north of the breeding colonies on St. Lawrence Island, representing the area used for foraging by seabirds in the focal locations (as documented in Will et al., 2020) and in general available to seabirds associated with colonies on the island’s north coast. Stations were spaced 20 nautical miles apart in 2017 and 2019 and spaced 30 nautical miles apart in 2018 from latitudes 63.3° to 65.0° North and longitudes −167.5° to −168.35° West. We provide a comparison of fish abundances measured in the selected area to those in the full survey region of the northern Bering Sea shelf, excluding stations north of 65° North and west of 167° West. A 30-min tow was successfully completed at a total of 30, 14 and 31 stations in 2017, 2018 and 2019, respectively; we report mean number of individuals caught and biomass by station to account for differences in survey years. All species captured were identified, weighed, and enumerated with select fish species chosen for random length measurements. Catch per unit effort at each station was calculated in terms of weight (kg) and number of fish, with effort calculated as the distance towed * net width (km²). Data were restricted by fish length or by mean individual fish weight per station, so that fish abundance reflected only fish that seabirds were known to feed on and physically capable of consuming. These fish were then categorized into three ecological groups (“benthic”, “forage”, or “gadids”) based on their habitat preferences and taxonomy (see Will et al., 2020 for details).

2.6. Harvest Survey

The peak of the die-off (Romano et al., 2020) overlapped with the period when St. Lawrence Islanders were harvesting murres and their eggs. With permission from the Native Village of Savoonga, in February 2019 we conducted a harvest survey to collect standardized information on the 2018 murre die-off from the community of Savoonga and to fill key data gaps on reproductive effort of murres during egg laying. We replicated the Alaska Migratory Bird Co-Management Council Harvest Assessment Program’s survey (including their survey methods, Naves, 2012) of all harvested birds and eggs in the region, and included an additional questionnaire specifically addressing the murre egg harvest. These additional questions were designed to obtain the community’s perception of the die-off and to determine whether it affected the number of murres that laid eggs. A list of all households was generated with the help of A. Waghiyi, M. Annogiyuk and P. Shoogukwruk and were randomly assigned household identification numbers. The list was split between two local surveyors (M. Annogiyuk, and P. Shoogukwruk) who recorded the household ID on survey response forms. The anonymity of responses was ensured by destroying the master list linking households to their identification numbers after sample collection. Heads of households could decline to participate in the survey, and surveyors would try to make contact with a household up to three times before listing it as a “no contact” household. All data are owned by the Native Village of Savoonga and are reported here with their permission.

2.7. Screening for Disease

In 2018, one short-tailed shearwater, two thick-billed murres and two crested auklet carcasses were shipped frozen to the USGS National Wildlife Health Center (NWHC) in Madison, Wisconsin. Necropsies were performed on the two murres and one of the auklets by a board-certified veterinary pathologist. Brain, trachea, esophagus, heart, lungs, liver, spleen, kidney, gonad, adrenal gland, and pectoral muscle were collected for histopathology from the three birds necropsied. Samples of tissues were collected for ancillary diagnostic testing as indicated by the necropsy findings. Cloacal and tracheal swabs were collected in viral transport media for avian influenza virus (AIV) and Newcastle Disease virus (NDV) testing by real-time reverse transcription polymerase chain reaction (rRT-PCR) using the current National Animal Health Laboratory Network procedures for the matrix gene in the respective viruses (SOP-AV-0068). PCR-positive samples were inoculated into day-8 embryonating chick embryos for attempted virus isolation (Ip et al., 2012). Potential pathogenicity of avian influenza viruses was determined by the intravenous pathogenicity index (IVPI) assay (OIE, 2015). Sequence characterization of avian influenza viruses were performed according to Lee et al., (2019). Sequence alignment using Muscle and phylogenetic tree construction using the Neighbor-Joining method with 1000 bootstraps were performed using MEGAX with default parameters (Kumar et al., 2018). Cloacal and oropharyngeal swabs of live birds (Suppl. Table 3) were placed into viral transport media during the 2019 field season and shipped frozen to the NWHC laboratory for AIV and NDV testing.

2.8. Statistical analysis

All analyses were completed in R versions 3.6.0 and 3.3.4 (R Core Team 2019). Corticosterone concentrations were standardized by mass, for feather samples (Will et al., 2019) or volume, for blood plasma samples, and log-transformed to meet assumptions of normality. We used analysis of variance to compare corticosterone concentrations, bird mass and fish abundance across years and, in some cases for corticosterone, between locations, followed by a Tukey HSD post-hoc analysis to identify between which variables significant differences occurred. Results were considered significant when p-value < 0.05. To calculate numbers of birds and eggs harvested from survey results we followed methods in Naves (2012) where the estimated harvest in Savoonga is the sum of the harvests reported by surveyed households multiplied by the proportion of households sampled per total households in the community. We report results as a percentage of change to avoid this work being used in a way that may harm subsistence practices. Estimates are available upon request to and approval by the Native Village of Savoonga.

3. Results

3.1. Colony Attendance

The average ratio of thick-billed to common murres observed on study plots in 2016–2018 was 7:1. Raw counts summed across the 11 study plots for thick-billed murres:common murres for the comparison dates in each year were: 2016 – 517:62, 2017 – 449:103, 2018 – 260:31. In 2019, after the die-off, the ratio was 3:1 (325:99).

3.2. Beach Surveys

In 2018 we encountered 420 bird carcasses (Fig 2), about 26 per km of surveyed beach, of which 61.1% (258) were thick-billed murres, 2.4% (10) were common murres, and 9.7% (41) were unidentified murre species (partial carcasses). Almost 26 thick-billed murre carcasses were found for every one common murre carcass. Short-tailed shearwaters (10.4%, 44) were the next most numerous taxon found, followed by other alcids (6.2%, 26; mostly pigeon guillemots, Cepphus columba, 11), larids (5.45%, 23), and northern fulmars (2.4%, 10, Fig 2).

Figure 2.

Beached bird surveys from 2018 (left) and 2019 (right). 420 bird carcasses were encountered in 2018, 66 carcasses were encountered in 2019. All birds were identified to the level of genus or species, here we show the larid and alcid family (except for murres) for simplification.

In 2019 we encountered 66 carcasses, ~5 per km of beach. Other alcids, primarily crested auklets (20), constituted the bulk of the carcasses (53.7%, 29), followed by thick-billed murres (24%, 13), larids (16.7%, 9), and other birds (14.8%, 8, Fig 2).

3.3. Exposure to Nutritional Stress

When compared across the four study years (2016–2019), concentrations of corticosterone in fall-grown primary feathers in all birds (those sampled alive and dead) were significantly higher (mean = 1.665 ± standard error 0.027 log[ng/g]) in the fall of 2017 than in any other year (2015: 1.397 ± 0.084 log[ng/g]; 2016: 1.493 ± 0.031 log[ng/g]; 2018: 1.512 ± 0.038 log[ng/g]; Tukey post-hoc HSD adjusted p < 0.01, for all 2017 comparisons, Fig 3). Concentrations of corticosterone in late-winter-grown throat feathers was significantly lower in 2019 (1.806 ± 0.042 log[ng/g]) than in 2016 (2.044 ± 0.056. log[ng/g]), 2017 (1.987 ± 0.035 log[ng/g]) and 2018 (2.001 ± 0.0426 log[ng/g]; p < 0.04, Fig 3). Baseline plasma corticosterone collected during late summer did not differ among the four study years (2016: 0.411 ± 0.047 log[ng/mL]; 2017: 0.526 ± 0.049 log[ng/mL]; 2018: 0.622 ± 0.082 log[ng/mL]; 2019: 0.523 ± 0.069 log[ng/mL]; F_1,87 = 2.07, p = 0.154, Fig 4). Across all years there was no correlation of corticosterone concentration between seasons (fall/primary feather to late winter/throat feather: F_1,55 = 3.33, p = 0.0734, late winter/throat feather to summer/plasma: F_1,55 = 0.42, p = 0.52, fall/primary feather to summer/plasma : F_1,55 = 1.45, p = 0.23).

Figure 3.

Corticosterone concentrations in fall grown primary feathers (left panel) and late-winter grown throat feathers (right panel) sampled from breeding thick-billed murres on St. Lawrence Island compared to historical concentrations from birds collected in the northern Bering Sea region. Fall 2017 and late winter 2018 values include samples from both live birds and beach carcasses. Boxplots indicate the data quartiles centered on the median, raw data are overlain as grey points, bold points are those that are 1.5 * the interquartile range below or above the middle 50% of the data (the value for quartile 1 and 3 respectively). Sample sizes are noted above each boxplot.

Figure 4.

Corticosterone concentrations in blood plasma sampled from breeding thick-billed murres on St. George Island in the southeastern Bering Sea (data reproduced here from Kokubun et al., 2018) and on St. Lawrence Island. Boxplots indicate the data quartiles centered on the median, raw data are overlain as grey points, bold points are those that are 1.5 * the interquartile range below or above the middle 50% of the data (the value for quartile 1 and 3 respectively). Sample sizes for this study are noted above each boxplot.

Thick-billed murres found dead in the summer of 2018 (1.684 ± 0.032 log[ng/g]) had significantly higher corticosterone concentrations in their fall (2017) grown primaries than did murres captured alive in 2018 (1.581 ± 0.02 log[ng/g]; Welch’s two-sample t-test: t_36.93 = −2.98, p = 0.005). There was no significant difference in corticosterone concentrations in the throat feathers grown in late winter between birds found dead and alive in summer 2018 (t_32.29 = −1.42, p = 0.16, Fig 5).

Figure 5.

A comparison of exposure to nutritional stress in thick-billed murres that were found dead (n = 32) and alive (n = 7) in the summer of 2018, during (left panel) the fall, when thick-billed murres are in the northern Bering and Chukchi Seas, and (right panel) late winter, when thick-billed murres are likely outside of the ice-covered Pacific Arctic. Boxplots indicate the data quartiles centered on the median, raw data are overlain as grey points, bold points are those that are 1.5 * the interquartile range below or above the middle 50% of the data (the value for quartile 1 and 3 respectively).

At the historical scale, concentrations of feather corticosterone in 2017 fall grown primaries were no different than concentrations in primaries from 1975, 1976, 1977, and 2018, but were significantly higher than all other years (1921, 1942, 1950, 1996, 1997, 2003; p < 0.01, Fig 3). Throat feathers grown in the late winter were significantly lower in 2019 than 2016, 2017, and 2018 (p < 0.04), but concentrations between all other years were similar (including 1958 and 1959; 1935 and 1972 were excluded from the analysis due to sample sizes of 1, Fig 3). In general, baseline plasma corticosterone concentrations were lower in St. Lawrence Island breeding thick-billed murres compared to thick-billed murres breeding on St. George Island (years 2003–2005, 2008–2011, 2013–2015; Colony F_1,616 = 10.61, p = 0.00119). Across both colonies baseline plasma corticosterone varied significantly over time (Year: F_13,616 = 10.39, p < 0.0001, Fig 4); however, 2018 concentrations were no different than any of the other previous years (comparisons including 2018; p = 0.51 for 2003, all other years p > 0.8)

Body mass of thick-billed murres captured at the colony was not different (F_3,94 = 1.25, p = 0.296) between 2016 (966.2 ± 13.75 g n = 28), 2017 (995.2. ± 15.18 g, n = 39), 2018 (989.3 ± 25.17 g, n = 7), and 2019 (1007.5 ± 13.16 g, n = 24).

3.4. Prey Abundance

Within the study area north of St. Lawrence Island, the abundance of small forage size fishes was no different among the years 2017–2019 (number of individuals: F_2,70 = 0.55, p = 0.58; biomass: F_2,70 = 0.22, p = 0.8; Fig 6). Small-sized benthic fish abundance by number of individuals and biomass in 2018 was the same as in 2017 and 2019 (Fig 6, number: F_2,70 = 0.84, p = 0.44; biomass: F_2,70 = 0.66, p = 0.52). Gadid abundance by number of individuals and biomass also did not change between 2017–2019 (gadid number: F_2,70 = 0.4, p = 0.67; gadid biomass: F_2,70 = 0.05, p = 0.96). Forage fish number and biomass of individuals did differ among the study years (forage fish number: F_2,70 = 3.33, p = 0.04; forage fish biomass: F_2,70 = 4.31, p = 0.02). The number of forage fish individuals was significantly lower in 2019 than 2017 (p = 0.05), but was no different between 2017 and 2018, or between 2018 and 2019 (p > 0.2). Forage fish biomass north of St. Lawrence Island was significantly lower in 2019 compared to 2017 and 2018 (p < 0.05), but did not differ between 2017 and 2018 (p = 0.95).

Figure 6.

Fish abundance as the mean catch per unit effort (CPUE) log-transformed number of individuals and mean CPUE log-transformed biomass in kg/km² for 2017–2019. Bottom-trawl survey results are displayed for the area north of St. Lawrence Island (top), and the greater Northern Bering Sea region (bottom).

In the larger northern Bering Sea survey region, bottom trawl surveys showed no change in small forage size fish abundance in 2017–2019 (number of individuals: F_2,253 = 1.58, p = 0.21; biomass: F_2,253 = 2.28., p = 0.12; Fig 6). Small-sized benthic fish abundance did not change (number: F_2,253 = 0.04., p = 0.96; biomass: F_2,253 = 1.14., p = 0.32), however gadid and forage fish abundance did differ between years. There were significantly more gadids, as measured by number of individuals, in 2017 than in 2018 (p = 0.03), however there were similar numbers between 2018 and 2019 (p = 0.06). The biomass of gadids was significantly higher in 2019 than 2017 or 2018 (p = 0.01 for both), and did not differ between 2017 and 2018 (Tp = 0.81). Forage fish numbers and biomass was significantly lower in 2019 compared to 2017 (p < 0.0005), but did not differ between 2017 and 2018 (by number p = 0.42, biomass p = 0.09) or between 2018 and 2019 (p = 0.2 for both number and biomass).

3.5. Harvest Survey

We obtained 88 complete surveys from 144 households in Savoonga (61% response rate). Six households declined to participate in the survey, and we attempted, but were unable, to make contact with 9 others. In total 71.5% of households had the opportunity to participate in the survey. Of the households that actively harvested murre birds or eggs between 2015–2018, 84% participated in the harvest of murres in 2018. Reasons for forgoing harvest were not exclusively due to there being few murres and eggs but also included working, taking care of children, health issues, and no transportation. Murre bird harvest in 2018 was one fifth (18.2%) of the mean 1993–2012 harvest, and three quarters (73.2%) the size of the 2012 harvest. Murre egg harvest in 2018 was one fifth (20.6%) of the mean 1993–2012 harvest, and half (47.6%) the size of the 2012 harvest. 56% of households said the egg harvest was less than it was in the past 5 years (2013–2018), while the remaining households observed no difference. Forty percent of households reported that the timing of the murre egg harvest was the same as in the last 5 years, and another 40% reported that the timing was later than in the last 5 years. Three harvesters said that they thought the murres may be sick, and described unusual behaviors such as swimming in circles near shore. One household noted that murres were observed laying eggs in late August.

3.6. Screening for Disease

Of the five bird carcasses submitted for necropsy in 2018, complete post-mortem examination was performed on three birds (two thick-billed murres found dying on the beach and one crested auklet, Suppl. Table 2). All three birds had no subcutaneous fat and showed evidence of atrophy of the pectoral musculature (emaciation), however one thick-billed murre’s mass was 950 g, within the ranges observed in this study during non-die-off years. The crested auklet (28938–004) also had puncture wounds indicative of trauma from predation. Major organs were unremarkable on histopathology. All cloacal and tracheal swab samples were negative for the matrix gene avian influenza rRT-PCR test with the exception of the cloacal swab from one of the thick-billed murres (28938–003, 950 g individual noted above, Suppl. Table 2). Molecular testing results from this swab were confirmed independently by two laboratories, and a H10N6 avian influenza virus was isolated by both laboratories when the samples were inoculated into embryonating chicken eggs. Whole genome sequencing of the isolate revealed that most (five) of the gene segments belong to typical North American wild bird low pathogenic avian influenza virus lineages, but three genes (HA, PB1 and PA) were of Eurasian lineage origin (Fig 7 and data not shown). None of the cloacal and oropharyngeal swab samples collected in 2019 were positive for AIV and NDV (Suppl. Table 3). None of the chickens inoculated with the H10N6 virus developed symptoms and there were no deaths (IVPI = 0.0).

Figure 7.

Phylogenetic analysis of the H10 (A) and N6 (B) RNA segments using the Neighbor-Joining method. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) are shown next to the branches. Representative viruses from the Eurasian lineage are denoted with a white circle symbol and from the North American lineage with a black circle symbol. The 2018 murre H10N6 virus is denoted with a red diamond symbol.

4. Discussion

We used colony-based observations and physiology to investigate possible causes of starvation observed in the 2018 seabird die-off in the northern Bering Sea. Thick-billed murres were most affected by the 2018 die-off in the vicinity of St. Lawrence Island–the ratio of thick-billed to common murres among dead birds was several times higher than expected based on our counts of both murres breeding on St. Lawrence Island, ratios of breeding thick-billed to common murres breeding in the broader continental regions of the Bering and southern Chukchi Seas (Romano et al., 2020), and ratios of thick-billed to common murres counted at sea in the Bering Sea shelf regions (Kuletz et al., 2019, 2020; Romano et al., 2020). Birds sent in for necropsy were determined to be emaciated. We examined potential causes of emaciation, including food shortage and disease. Corticosterone concentrations in fall grown primary feathers indicate that food shortages occurred in the fall prior to the mass mortality event. Birds found dead in the summer of 2018 had experienced higher nutritional stress in the fall of 2017. However, we found no evidence that birds were exposed to food shortages on their wintering grounds prior to returning north in the spring. Baseline plasma corticosterone concentrations in St. Lawrence Island thick-billed murres during the summer were comparable to years of adequate food availability in thick-billed murres breeding on St. George Island in the southeastern Bering Sea (Kokubun et al., 2018). The abundance of small forage sized fish north of St. Lawrence Island was no different in 2018 compared to 2017 and 2019.

We did not find support for a severe food shortage being the sole cause of the thick-billed murre mortality event. Thick-billed murres breeding on St. Lawrence Island rely on small forage sized fishes. Both the number of individuals, and the biomass of fish caught in bottom trawl surveys north of St. Lawrence Island, and the larger northern Bering Sea region, did not differ in 2018 compared to 2017 and 2019, years in which no disproportionately high mortality of thick-billed murres occurred. Will et al. (2020) demonstrated that during 2016–2019 thick-billed murre diet did not change from early to late summer (reflected in stable isotope values) and that thick-billed murres appear to rely heavily on benthic fishes (also supported by Springer et al., 1987; Tsukamoto, 2019). These small benthic fish would likely have been available to birds upon their arrival in the spring, as there is some evidence that related species tend to have relatively restricted movements (< 150 km, Horton, 1989; Piatt, 2002; Meyer Ottesen et al., 2018; Landry et al., 2019). Since thick-billed murres are successful at foraging on a wide range of prey species, we would expect that if their benthic prey were limiting they would have switched to consuming small forage size pelagic fish species, preferred by common murres, or euphausiids. If thick-billed murres had exhausted all of their food options, small benthic fish, pelagic forage fish, and euphausiids, and were starving solely because of lack of prey then we would expect to have seen the die-off affect proportionate numbers of thick-billed and common murres and/or planktivorous seabirds such as crested auklets. This was not the case. In fact, we observed that, in contrast to thick-billed murres, both common murres and crested auklets experienced high nutritional stress in 2018 (Will et al., 2020) but did not die in proportionate numbers to thick-billed murres. The divergent responses to conditions between common and thick-billed murres indicate that benthic prey was available to thick-billed murres throughout the spring and summer of 2018, and that they were not food limited compared to 2017 or 2019 when no die-offs occurred.

Egg laying in 2018, inferred from the egg harvest, was low, and the timing of laying may have been somewhat delayed, although harvest survey responses on the latter were split and inconclusive. Until murres have committed to incubating they are known to have sporadic colony attendance, largely driven by the presence or absence of food (Gaston and Nettleship, 1982). If the food shortage was severe enough to kill large numbers of individuals, thick-billed murres, as long-lived organisms would most likely have abandoned the breeding colony altogether (Lancaster et al., 2007; Piatt et al., 2020), as the high nutritional stress experienced during a severe food shortage would down-regulate hormones necessary for the formation of eggs (Kamel and Kubajak, 1987). Thus, the fact that some birds did attempt to breed and at a relatively normal time, while others were still dying, further indicates that food was not severely limiting, and that perhaps some other factor was differentially affecting the ability of murres to capture prey. In addition, almost half of Savoonga households (44%) reported no difference in their harvest numbers in 2018 compared to the past 5 years, indicating that interpretation of the harvest conditions was somewhat variable and enough birds laid eggs to provide some families with an adequate take.

We also considered the potential role disease may have played in the starvation experienced by thick-billed murres in 2018. Disease has been an increasing factor in mass mortality events in birds worldwide (Fey et al., 2015). Avian influenza has been documented to occur in the St. Lawrence Island population of murres at a background rate of 2–4% (common and thick-billed murres respectively, Ip et al., 2008). The novel strain of avian influenza detected in 2018 (this study) and some of the strains previously found in seabirds on St. Lawrence Island contain Eurasian gene segments (Ramey et al., 2010), which birds may be exposed to during their winter migration; in 2016 St. Lawrence Island breeding thick-billed murres were documented overwintering in the Sea of Japan and Sea of Okhotsk (Takahashi et al., 2020). Experimental work in migratory red knots indicates that prior to migration birds may be particularly susceptible to contracting avian influenza, and can carry productive viral loads without showing clinical signs of the disease (Reperant et al., 2011). Thick-billed murres, therefore, may have carried the virus during migration from their wintering grounds (e.g. Lebarbenchon et al., 2015), it may have flared up in birds whose systems had been taxed by that pre-breeding migration, and led quickly to starvation in weakened murres unable to meet their daily energy requirements (Elliott and Gaston, 2014; Piatt et al., 2020) a multiple-stressor scenario (Owen et al., 2012; Sebastiano et al., 2016) that is increasingly associated with die-offs across animal taxa (Fey et al., 2015). Our results provide some support for this connection; birds that died in the summer of 2018, had higher concentrations of corticosterone in their fall (2017) grown first primaries, and may have been pre-disposed to experience costs associated with the return migration in the spring. Unfortunately, except for the two moribund thick-billed murres we collected, we lack fresh carcasses from the die-off that could be examined for clinical signs of disease. Experimental inoculation with the murre H10N6 virus in chickens resulted in no morbidity or mortality, indicating that this H10N6 is a low pathogenic avian influenza virus, at least as defined by the laboratory model – the domestic chicken. Infection of wild birds with low pathogenic avian influenza viruses are thought to be inapparent or mild (Kuiken, 2013), but may result in reduced forage rates and delayed migratory departure (van Gils et al., 2007), or increased susceptibility to other infections (Suttie et al., 2018). Recovery from these low pathogenic strains can be slow, and effects may persist into the next season (Hoye et al., 2016). Even so, our findings of a low pathogenic strain of avian influenza in a single thick-billed murre do not provide unequivocal support for the hypothesis that thick-billed murres suffered from a disease outbreak (but see Zohari et al., 2014).

Nevertheless, several characteristics of the H10N6 virus isolated from the thick-billed murre are notable. The H10N6 virus detected in 2018 was different than previously documented viruses, including H10N6 subtypes previously described in St. Lawrence Island birds (Ramey et al., 2010, see Fig 6). The amino acid sequence at the protease cleavage site of the hemagglutinin protein is one of the factors that contributes to virulence of avian influenza viruses (Senne et al., 1996), and the deduced amino acid sequence at the murre H10N6 hemagglutinin cleavage site (PELMQGR/G) differs at only one residue from that of other highly pathogenic H10 subtype avian influenza viruses (PEIMQGR/G, Kim et al., 2012). Furthermore, several additional studies have demonstrated that H10 subtype avian influenza viruses, including an H10N6 subtype associated with high mortality in laboratory mice, have the potential to replicate in both birds and mammals (Englund, 2000; Arzey et al., 2012; El-Shesheny et al., 2018; Guan et al., 2019). Collectively, these findings indicate that further characterization of the properties of the H10N6 virus isolated from the thick-billed murre is warranted, and the discovery of this virus from a moribund individual supports the utility of mortality-based surveillance in informing awareness and timely reporting of pathogen risk to communities that rely on migratory birds to maintain cultural practices and food security. This surveillance is also critical in monitoring what viral subtypes are circulating among migratory birds of North America (Takahashi et al., 2020).

Food shortages detected in fall 2017 using corticosterone concentrations in fall-grown primary feathers were comparable in severity to food shortages in 1975–1977. The similarity in nutritional stress exposure between the falls of 2017 and 2018, and the falls of 1975–1977 provide bird-based evidence that a reorganization of the Pacific Arctic food web may be underway. Thick-billed murres molt their primaries in the fall, after the breeding season when murres breeding on St. Lawrence Island may be in the southern Chukchi Sea (Kuletz et al. 2015, Takahashi et al., 2020). Differences in habitat use may explain why thick-billed murres showed nutritional stress in the fall, but not in the summer. Alternatively, but not mutually exclusive, they may rely on different prey in the fall.

Our investigation into what caused the mass mortality of thick-billed murres in 2018 shows that a severe food shortage was unlikely the cause of the starvation observed on St. Lawrence Island. Abundance of forage size fishes north of St. Lawrence Island was not different in summer 2018 compared to the summers of 2017 and 2019, and were likely indicative of what was available to murres in late spring and early summer. These findings are not necessarily applicable to the larger area affected by the 2018 murre die-off. Limited data on the specific species of dead murres (Romano et al. 2020), an absence of the nutritional status of murres at other colonies, and differences in overwinter migratory routes (Hatch et al.; 2000, Takahashi et al., 2020), prey availability and oceanography make it possible that starvation in murres elsewhere was caused by a different balance of factors. Of eight additional salvaged thick-billed murres sent in for necropsy from around the northern Bering Sea and Bering Strait region (time since death not known), none yielded positive detection of avian influenza. RNA viruses such as avian influenza rapidly degrade in carcasses, and require relatively prompt freezing upon the point of death to optimize virus preservation (Munster et al., 1999; Wille et al., 2018). On St. Lawrence Island, we found that thick-billed murres were most affected by the die-off and that the cause of mortality was transient; neither food shortages nor disease were evident in the population by late July, highlighting the need for immediate collection of moribund seabirds to effectively detect non-food related causes of mass mortality events.

The impact of the die-offs on the local population of murres was noticeable. In 2019 our counts of thick-billed murres dropped by almost 25% from those observed in 2017 compared to common murres whose numbers in 2019 were comparable to those in 2017. Having occurred during the breeding season, the impact of the die-off may be more localized, compared to wintertime die-offs. In the winter of 2015–2016 an unprecedented die-off of common murres occurred in the Gulf of Alaska (Piatt et al., 2020). It is possible that common murres breeding on St. Lawrence Island were affected by this die-off (Hatch et al., 2000; Takahashi et al., 2020). In the late 1980s, breeding common and thick-billed murres were observed to occur in equal proportions on St. Lawrence Island (Piatt et al., 1988), and even more recent (2007–2015) at-sea summer/fall surveys indicated that thick-billed murres were only slightly more numerous in recent decades (1.5:1, Kuletz et al., 2019), compared to the 7:1 ratio we observed in 2016–2018. These potential population-level effects of winter and summertime die-offs may have lasting effects on communities like Savoonga that use murres as a food resource, and which may benefit from more rapid responses to determining the cause of mass mortality events.