Migratory culture, population structure and stock identity in North Pacific beluga whales (Delphinapterus leucas)

Abstract

The annual return of beluga whales, Delphinapterus leucas, to traditional seasonal locations across the Arctic may involve migratory culture, while the convergence of discrete summering aggregations on common wintering grounds may facilitate outbreeding. Natal philopatry and cultural inheritance, however, has been difficult to assess as earlier studies were of too short a duration, while genetic analyses of breeding patterns, especially across the beluga's Pacific range, have been hampered by inadequate sampling and sparse information on wintering areas. Using a much expanded sample and genetic marker set comprising 1,647 whales, spanning more than two decades and encompassing all major coastal summering aggregations in the Pacific Ocean, we found evolutionary-level divergence among three geographic regions: the Gulf of Alaska, the Bering-Chukchi-Beaufort Seas, and the Sea of Okhotsk (Φst = 0.11-0.32, Rst = 0.09-0.13), and likely demographic independence of (Fst-mtDNA = 0.02-0.66), and in many cases limited gene flow (Fst-nDNA = 0.0-0.02; K = 5-6) among, summering groups within regions. Assignment tests identified few immigrants within summering aggregations, linked migrating groups to specific summering areas, and found that some migratory corridors comprise whales from multiple subpopulations (PBAYES = 0.31:0.69). Further, dispersal is male-biased and substantial numbers of closely related whales congregate together at coastal summering areas. Stable patterns of heterogeneity between areas and consistently high proportions (~20%) of close kin (including parent-offspring) sampled up to 20 years apart within areas (G = 0.2-2.9, p>0.5) is the first direct evidence of natal philopatry to migration destinations in belugas. Using recent satellite telemetry findings on belugas we found that the spatial proximity of winter ranges has a greater influence on the degree of both individual and genetic exchange than summer ranges (rwinter-Fst-mtDNA = 0.9, rsummer-Fst-nDNA = 0.1). These findings indicate widespread natal philopatry to summering aggregation and entire migratory circuits, and provide compelling evidence that migratory culture and kinship helps maintain demographically discrete beluga stocks that can overlap in time and space.

Fig 1. Distribution (light blue) of beluga whales, Delphinapterus leucas, in the North Pacific Ocean.

The ten major nearshore concentration areas during the summer months are highlighted (dark blue). These areas along with a small resident group of beluga whales in the Gulf of Alaska are numbered according to Table 1.

Fig 2. Migration routes and sampling locations of beluga whales from the major summer concentration areas in the Bering, Chukchi and Beaufort Seas and from the Gulf of Alaska.

Summering and wintering areas, and migration routes were inferred from a combination of satellite telemetry, aerial and shore based sightings, and Traditional Ecological Knowledge (TEK). Sampling sites are indicated by yellow circles and in the case of migration by triangles.

Fig 3. Summary plots generated in Clumpak of model-based cluster analysis of population structure in Pacific beluga whales using STRUCTURE 2.3.4.

The major modes for K = 4 to 6 (based on five separate runs for each value of K) are presented for the analysis using prior sample group information and no admixture which revealed K = 5 clusters as the most likely (see panel 2). However, in a number of analyses K = 6 was the most or second-most likely resulting in the separation of Anadyr into a discrete cluster (see panel 3). Each genotyped individual is represented by a vertical line with estimated membership, Q, in each cluster denoted by different colors. The analysis was based on using all individuals (n = 1032) scored at 6 or more loci (n_loci≥6).

Fig 4

Mantel tests of the correlation between genetic differentiation (F_st) and geographic distance among summering and wintering grounds of beluga whales in the Bering, Chukchi and Beaufort Seas, for both (A) mitochondrial DNA and (B) microsatellite markers. Test p values are based on 10,000 permutations of the distance data.

Fig 5. The likely population of origin of beluga whales on spring migration sampled at four locations in the Bering, Chukchi and Beaufort Seas.

Individual assignments are represented as the relative height of stacked bars to either of two baseline populations, the eastern Beaufort Sea (blue) or eastern Chukchi Sea (red) for mtDNA (dark shading) and nDNA (light shading). A: Maximun Likelihood assignments in Whichrun. B: Baysian assignments using prior sampling information and admixture models in Structure 2.3.4. C: Mixed-stock assignments in Bayes. Only individuals with complete mtDNA-nDNA profiles are shown. See Table 4 for more details.

Fig 6. The probability distribution of population proportions of groups of beluga whales sampled on northbound migration in spring.

Stock-mixture analysis was conducted in Bayes with the eastern Beaufort Sea (blue) and the eastern Chukchi Sea (red) as baseline populations and the migrating groups as the potential ‘mixtures’. The ordinate axis indicates the number of runs.

Fig 7. The proportion of pairwise genealogical relationships estimated for beluga whales sampled within and between years across two decades near Kasegaluk Lagoon, Alaska.

Maximum likelihood estimates of four relationship categories were estimated from genotypic data using the program ml-relate. The stacked bars represent the proportions of distantly/unrelated individuals to closely related individuals (i.e., parent-offspring, full-sib and half-sib or equivalent) for a subset of the 20-year data set comprising the first three years (1988, 1993, 1994) and the last three years (2005, 2006, 2007).

Fig 8. Test of differences in mean relatedness (r¯) among beluga whales within a single year compared to r¯ between whales from two different years for the same summering ground using coancestry.

The graph depicts results for Kasegaluk Lagoon 1988 compared to 1988–2007 using the ML estimator TrioML of Wang (2007). If the observed difference (black line) falls outside the 90% (dotted lines), 95% (dashed lines), and 99% (green solid lines) confidence intervals from the bootstrap analysis distribution the difference is adjudged to be significant.