U.S. Dept Commerce/NOAA/NMFS/NWFSC/Publications

NOAA-NWFSC Tech Memo-25: Status Review of Pink Salmon from Washington, Oregon, and California Coast

Genetic Information

Differences Between Even- and Odd-year Pink Salmon

Because pink salmon mature and spawn on a strict 2-year cycle, genetic isolation between odd- and even-year spawners is nearly complete, and several genetic and biological variables differ between them. The electrophoretic analysis of enzymatic proteins has been used extensively to measure the genetic differences between odd- and even-year fish and to resolve genetic structure among populations within these groups. One approach for detecting reproductive isolation is to compare frequencies of protein variants (allozymes) among samples. The finding of significant frequency differences between groups can be taken as evidence of reproductive isolation. Another approach to identifying reproductively isolated groups is to estimate genetic distances between samples and to analyze these distances with a clustering algorithm, such as the unweighted pair group method with averages (UPGMA; Sneath and Sokal 1973) or with an ordination technique such as multidimensional scaling analysis (Lessa 1990). When the geographic distribution of genetic variability is continuous and not hierarchical or disjunct, such as in a clinal or reticular pattern, multidimensional scaling is more appropriate than agglomerative clustering (Lessa 1990). Multidimensional scaling is a nonmetric ordination technique that depicts genetic relationships among populations in two or three dimensions and can reduce the distortion that may exist in phenograms because populations are represented in fewer dimensions. When genetic relationships among populations reflect nonhierarchical or semihierarchical geographic variation, multidimensional scaling diagrams are often a more effective means of showing these relationships (Lessa 1990).

Geneticists have used several genetic distance measures (e.g., Cavalli-Sforza and Edwards 1967, Rogers 1972, Nei 1978) to study the population structure of pink salmon as well as other salmonids. A considerable literature has developed on the pros and cons of these measures. For example, an attractive feature of the Rogers and Cavalli-Sforza and Edwards distances is that they satisfy the triangle inequality--that is, given three populations (A, B, C), the sum of the distances from A to B and from A to C will always be greater than the distance from A to C. On the other hand, neither of these distance measures employs a correction for sample size, so the distances are biased upwards, especially for small sample sizes. In contrast, Nei s distance is unbiased, but does not always satisfy the triangle inequality. Another important consideration is that both Nei s and Rogers distance measures can be affected by different levels of heterozygosity between populations, whereas Cavalli-Sforza and Edwards measure is not. Discussions of these and other features of genetic distances appear in Nei (1987), Hillis and Moritz (1990), and Rogers (1991). Unfortunately, most of this discussion has focused on the merits of the various measures for phylogenetic reconstruction among species and higher taxa. No one has rigorously or quantitatively evaluated the performances of these distances in assessing the genetic population structures of species like salmon, which typically are separated by relatively small genetic distances.

Since it is unclear which distance measure is best in any given application, we analyzed each set of data with more than one method to identify results that may not be robust. Nei s unbiased genetic distance has been used in several studies of pink salmon, but we computed all three distance measures for most data sets. In most cases, the different genetic distance measures yielded results that were highly correlated. For simplicity we report only results for Rogers and Cavalli-Sforza and Edwards distance measures. Both distance measures range from 0.0 (identity) to 1.0 (complete dissimilarity). Cases in which the results differed substantially among measures are identified in the text.

For many polymorphic enzyme-encoding loci, strong allozyme frequency differences have been reported between even- and odd-year broodlines spawning at the same localities in Alaska (Aspinwall 1974, Johnson 1979, McGregor 1982), Canada (Beacham et al. 1985), and Russia (Salmenkova et al. 1981, Altukhov et al. 1983, Kartavtsev 1991). Shaklee and Varnavskaya (1994) reported a large genetic difference between even- and odd-year pink salmon from the Snohomish River, the only North American locality south of British Columbia that supports a spawning population of even-year pink salmon. The average Nei s genetic distance (Nei 1978) between even- and odd-year pink salmon in British Columbia was 0.018 (Beacham et al. 1988). However, since this estimate was based only on polymorphic loci, it is likely to overestimate the true genetic distance between the two broodlines. The actual distance, which includes monomorphic loci in its estimate, is probably not more than half this value (or about 0.009). This distance is typical of conspecific populations of other animals (Thorpe 1982), and represents an upper bound to the largest distance expected to exist between populations within broodlines.

In addition to allozyme frequency differences, other genetic differences have been reported between even- and odd-year spawners. Gorshkova (1983) found that Kamchatka odd- year pink salmon show a chromosomal polymorphism of presumably acrocentric fusions, in which the diploid number is 53 or 54 chromosomes, whereas even-year spawners consistently had only 52 chromosomes.

At Auke Creek in southeastern Alaska, Gharrett and Smoker (1991) made crosses between even- and odd-year spawners to search for possible outbreeding depression in their offspring. Cyropreserved sperm from even-year males was used to fertilize eggs from odd-year females the following year. In the first generation, the number of returning first-generation hybrids and their average date of return were not significantly different from the respective numbers and return times of control fish of the same age, but variability among the hybrids in morphological characters was greater than that among control fish. Second-generation hybrids showed a reduced return rate and greater bilateral asymmetry in meristic characters, which were interpreted as possible genetic effects of outbreeding depression in hybrids between the two broodlines.

Even-year Pink Salmon: Genetic Variability Among Regions

Even-year adult pink salmon spawn throughout much of the species range but tend to increase in abundance with latitude (Heard 1991). In southern British Columbia and Washington, even-year spawners are less abundant than odd-year spawners, but they are as abundant or more abundant than odd-year spawners in parts of northern British Columbia and western Alaska. Even-year spawners are also abundant in Asia and tend to outnumber odd-year spawners in northern areas. Most genetic studies of even-year spawners have been made in Alaska and British Columbia; the information available for Asian even-year pink salmon is more limited. Several studies indicate that the degree of genetic differentiation among populations of even-year pink salmon is consistently lower than the degree of genetic differentiation between even- and odd-year spawners (McGregor 1983, Beacham et al. 1988, Gharrett et al. 1988, Shaklee and Varnavskaya 1994, Zhivotovsky et al. 1994).

No single study has included samples collected over the entire range of even-year spawners. Zhivotovsky et al. (1994) analyzed allelic frequencies for 20 loci (sAAT-3*, ADA-2*, mAH-3*, mAH-4*, CK-A1*, CK-A2*, G3PDH-1*, GPI-B1*, GPI-B2*, GPI-A*, LDH-A1*, LDH-A2*, LDH-B2*, LDH-C*, sMDH-A1,2*, sMDH-B1*, sMDH-B2*, mMEP-1*, MPI*, PEPD-1*, PEPD-2*, and sSOD-1*) in samples from southeastern Alaska (McGregor 1982, Lane et al. 1990), northwestern Alaska (Gharrett et al. 1988), and Hokkaido in Japan (Noll et al. 1994). These samples encompassed about three quarters of the geographic range of even-year pink salmon. Zhivotovsky et al. (1994) conducted a gene diversity analysis, which partitions the total genetic diversity observed in a set of samples into its regional components, at five hierarchical levels. The results indicated that 2.7% of the diversity was due to genetic differences between odd- and even-year spawners, and 2.3% was due to differences between Asian and North American samples. About 1.8% of the total diversity was due to regional differences between southeastern Alaska and western Alaska (including the Aleutian Islands). About 2% of the diversity was due to within-region variability, including geographic and run-timing differences in the same river system.

In samples from the Aleutian Islands, northwestern Alaska and Kodiak Island, Gharrett et al. (1988) examined 29 enzymatic loci (21 of which were polymorphic) and found the greatest regional genetic differentiation between the group of Aleutian Island-northwestern Alaska samples and a single sample from southcentral Alaska. Gharrett et al. (1988) then combined allelic frequencies for sMDH-B1*,sMDH-B2*; PGM-2*, ME-1*, G3PDH-1*, and PGDH* for Alaskan samples with those for samples from Sakhalin Island in Russia (Salmenkova and Omel chenko 1982, Altukhov et al. 1983). A maximum-likelihood tree based on these frequencies indicated that the Russian samples were most closely related to Alaskan samples from Norton Sound and the Aleutian Islands and more distantly related to samples from Bristol Bay. The sample from southcentral Alaska was most distantly related to the Russian samples. Although these results are consistent with those obtained by Zhivotovsky et al. (1994), it is not clear whether the major genetic discontinuity between eastern and western even-year spawners occurs between Asia and North America across the Bering Sea or across the Alaska Peninsula.

Several allozyme studies have been made of North American even-year spawning populations of pink salmon, but no study has included samples from the entire range of even-year spawners. Populations from northwestern Alaska appear to be most closely allied with Aleutian Island populations (Gharrett et al. 1988), but these authors did not address the relationship between these populations and those farther south. In British Columbia, Beacham et al. (1988) detected three regional groups: 1) populations on the Queen Charlotte Islands (these populations are distinct from all other British Columbia populations), 2) populations in the Skeena River and farther north, and 3) populations in central and southern British Columbia, including Vancouver Island.

An important question in evaluating the status of the Snohomish River even-year population is resolution of its origin. It may be a natural population or it may have originated from eggs translocated from Alaska or British Columbia (Table 4). To address this issue, we analyzed a set of allelic frequencies consisting of a single Snohomish River sample (J. Shaklee - footnote 22) and 34 samples from British Columbia (Beacham et al. 1988). This group of samples had 15 polymorphic loci in common, 9 of which showed common allelic frequencies of 0.95 or less in at least 1 sample. Frequencies for some loci were missing for Cluxewe, Glendale, Kemano, and Puntledge Rivers, so we substituted the average of frequencies in the two nearest populations for these loci. Other loci in these samples did not show strong allelic frequency differences between nearby populations. For odd-year spawning pink salmon, Shaklee et al. (1991) found a consistent difference in the scoring of two loci, sAAT-3* and PGDH*, between his laboratory and the laboratory that produced the data presented by Beacham et al. (1988) (see White and Shaklee 1991 for a discussion of this issue). Beacham et al. (1988) reported a 0.105 higher average frequency for the common allele of sAAT-3* and a 0.060 higher average frequency for the common allele of PGDH* in his samples than was found by J. Shaklee in samples taken from the same localities in different years. We therefore adjusted allelic frequencies for these two loci in the 34 samples from British Columbia. The adjustment was based on a comparison of allelic frequencies in 13 populations of odd-year pink salmon reported in Beacham et al. (1988) and later studied by Shaklee et al. (1991).

This basic set of data was then combined for analysis with data from Nickerson (1979) for Prince William Sound, and McGregor (1983), Gharrett et al. (1988), and unpublished data provided to us by A. J. Gharrett (footnote 23) for western, central, and southeastern Alaska, and the Aleutian Islands (Table 6). We also examined allelic frequencies for Kodiak Island (Johnson 1979), but as only four polymorphic loci were in common with data from British Columbia and Washington, we did not use these results to draw conclusions. We hypothesized that if the Snohomish River even-year spawners descended from eggs transplanted from Alaska or British Columbia earlier in this century, they may show genetic affinities to present-day Alaska or British Columbia even-year populations.


Table 6. Samples used in analyses of even-year pink salmon allozyme variability. Locality numbers appear in Figures 6-13.
Locality
Source
No. Name Region of dataa
Washington
1. Snohomish River Washington 1
British Columbia
2. Read River Southern BC 2
3. Wortley River Southern BC 2
4. Wakeman River Southern BC 2
5. Kakweiken River Southern BC 2
6. Glendale Riverb Southern BC 2
7. Waukwaas River Vancouver Island 2
8. Beer River Vancouver Island 2
9. Adam River Vancouver Island 2
10. Keogh River Vancouver Island 2
11. Puntledge Riverb Vancouver Island 2
12. Cluxewe Riverb Vancouver Island 2
13. Quinsam River Vancouver Island 2
14. Koeye River Central BC 2
15. Kairnet River Central BC 2
16. Neekas River Central BC 2
17. Atnarko River Central BC 2
18. Kitimat River Central BC 2
19. Kemano Riverb Central BC 2
20. Clyak River Central BC 2
21. Quaal River Central BC 2
22. Khutzeymateen River Central BC 2
23. Nakina River Northern BC 2
24. Kitwanga River (Skeena River) Northern BC 2
25. Babine River (Skeena River) Northern BC 2
26. Lakelse River (Skeena River) Northern BC 2
27. Kwiramass River Northern BC 2
28. Copper River Queen Charlotte Islands 2
29. Pallant River Queen Charlotte Islands 2
30. Windy Bay Queen Charlotte Islands 2
31. Yakoun River Queen Charlotte Islands 2
32. Deena River Queen Charlotte Islands 2
33. Naden River Queen Charlotte Islands 2
34. Security River Queen Charlotte Islands 2
Alaska
35. Herring Cove Creek (pooled) Southeastern Alaska 3
36. Porcupine Creek (pooled) Southeastern Alaska 3
37. Sashin Creek (pooled) Southeastern Alaska 3
38. Lover s Cove Creek (pooled) Southeastern Alaska 3
39. Fish Creek (pooled) Southeastern Alaska 4
40. Peterson Creek-mainland (pooled) Southeastern Alaska 3
41. Auke Creek (pooled) Southeastern Alaska 3
42. Peterson Creek-island (pooled) Southeastern Alaska 3
43. Rocky Creek (intertidal) Prince William Sound 5
44. Rocky Creek (upper) Prince William Sound 5
45. Constantine Creek (intertidal) Prince William Sound 5
46. Constantine Creek (upper) Prince William Sound 5
47. Zillesenof Creek (intertidal) Prince William Sound 5
48. Hartney Creek (intertidal) Prince William Sound 5
49. Humpback Creek (intertidal) Prince William Sound 5
50. Koppen Creek (intertidal) Prince William Sound 5
51. Koppen Creek (upper) Prince William Sound 5
52. Olsen Creek (intertidal) Prince William Sound 5
53. Olsen Creek (upper) Prince William Sound 5
54. Lagoon Creek (intertidal) Prince William Sound 5
55. Lagoon Creek (upper) Prince William Sound 5
56. Millard Creek (intertidal) Prince William Sound 5
57. Duck River (intertidal) Prince William Sound 5
58. Cannery Creek (intertidal) Prince William Sound 5
59. Swanson Creek (intertidal) Prince William Sound 5
60. Mink Creek (intertidal) Prince William Sound 5
61. Mink Creek (upper) Prince William Sound 5
62. Erb Creek (intertidal) Prince William Sound 5
63. Erb Creek (upper) Prince William Sound 5
64. Larson Creek (intertidal) Prince William Sound 5
65. Kenai River Cook Inlet 4
66. Susitna River Cook Inlet 4
67. Kodiak Island (pooled) Kodiak Island 4,7
68. Unalaska Island Aleutian Islands 7
69. Umnak Island Aleutian Islands 7
70. Blue Fox Bay, Atka Island Aleutian Islands 7
71. Korovin Bay, Atka Island Aleutian Islands 7
72. Adak Island Aleutian Islands 7
73. Tanaga Island Aleutian Islands 7
74. Semisopochnoi (1,2) Aleutian Islands 7
75. North Kiska Island Aleutian Islands 7
76. Attu Island Aleutian Islands 7
77. Naknek River Bristol Bay 3
78. Nushagak River Bristol Bay 3
79. Kwiniuk River Norton Sound 3
80. Nome River Norton Sound 3
a Key to data sources:
1. J. Shaklee, unpubl. data. Washington Dep. Fish and Wildlife, P.O. Box 43151, Olympia, WA 98501.
2. Beacham et al. (1988).
3. McGregor (1983).
4. A. J. Gharrett, unpubl. data. Juneau Center, School of Fisheries and Ocean Sciences, Univ. of Alaska, Fairbanks, 11120 Glacier Highway, Juneau, AK 99801.
5. Nickerson (1979).
6. Johnson (1979).
7. Gharrett et al. (1988).

b Missing data estimated by averages of closest localities.


In the first analysis, allelic frequencies for populations in Prince William Sound (Nickerson 1979) were included with frequencies for populations in Washington and British Columbia and used to estimate Rogers (1972) genetic distances based on 10 polymorphic loci (sAAT-3*, GPI-A*, GPI-B1*, G3PDH-1*, LDH-B2*, sMDH-A1*, sMDH-B1*, mMEP-1*, PGDH*, and PGM-2*). The UPGMA phenogram (Fig. 6) showed two distinct, nonoverlapping clusters: one including Prince William Sound samples (localities 43-64), and the other including British Columbia samples (localities 2-34). The Snohomish River sample (locality 1) fell within the cluster of British Columbia samples. Multidimensional scaling analysis (NTSYS-pc; Rohlf 1993) of these distances (Fig. 7) showed a similar geographical arrangement of the Prince William Sound and British Columbia samples, with the Snohomish River sample appearing at the edge of the British Columbia cluster on the opposite side of the cluster space from the Prince William Sound samples.

The second analysis included Snohomish River (locality 1), British Columbia (localities 2-34), and several localities in southeastern (localities 35-42) and southcentral Alaska (localities 65-67; McGregor 1983, A. J. Gharrett - footnote 24), the Aleutian Islands (localities 68-76; Gharrett et al. 1988), and western Alaska (localities 77-80; McGregor 1983). A UPGMA phenogram (Fig. 8) of Rogers genetic distances based on 10 polymorphic loci in common among the studies (sAAT-3*, ADA-2*, CK-A1*, GPI-B1*, G3PDH-1*, sMDH-A1*, sMDH-B1*, mMEP-1*, PGDH*, and PGM-2*) showed 2 major clusters. One cluster included samples from the Aleutian Islands, Kodiak Island and western Alaska, and the other included samples from southeastern Alaska and British Columbia. The sample from the Snohomish River was located within the cluster of British Columbia samples. Multidimensional scaling of Rogers genetic distances showed close agreement between the geographic locations of the samples and their general positions in the clusters apparent in the multidimensional scaling (Fig. 9). Samples from western Alaska and the Aleutian Islands were located at one end of the multidimensional space; those from southeastern Alaska were placed in an intermediate position, and those from British Columbia were at the other end of the multidimensional space. The Snohomish River sample occupied a position at the extreme end of the multidimensional space, opposite the samples from Alaska.

The results of these analyses demonstrate that the even-year Snohomish River sample is genetically more closely related to even-year populations from British Columbia than to those from Alaska. If this sample is representative of Snohomish River even-year pink salmon, these analyses lend strong support to an argument that this population did not arise from an Alaskan transplant.


Figure 6
Figure. 6. UPGMA phenogram of Rogers' (1972) genetic distances, based on 10 polymorphic loci (see text), between samples of even-year pink salmon collected in Washington (J. Shaklee, Washington Department of Fish and Wildlife, P.O. Box 43151, Olympia, WA 98501, unpubl. data), British Columbia (Beacham et al. 1988), and Prince William Sound, Alaska (Nickerson 1979). Locality numbers are given in Table 6.


Figure 7
Figure 7. Multidimensional scaling and minimum spanning tree (a tree connecting nearest genetic neighbors) of Rogers' genetics distances, based on 10 polymorphic loci (see text), between samples of even-year pink salmon samples collected in Washington, British Columbia, and Prince William Sound, Alaska. Locality numbers and sources of data as in Figure 6.


Figure 8
Figure 8. UPGMA phenogram of Rogers' genetic distances, based on 10 polymorphic loci (see text), between samples of even-year pink salmon from central and western Alaska (McGregor 1983), the Aleutian Islands, (Gharrett et al. 1988), central and southeastern Alaska (McGregor 1983; A.J. Gharrett, Juneau Center, School of Fisheries and Ocean Sciences, University of Alaska-Fairbanks, 11120 Glacier Highway, Juneau, AK 99801, unpubl. data), British Columbia (Beacham et al. 1988), and the Snohomish River, Washington (J. Shaklee, Washington Department of Fish and Wildlife, P.O. Box 43151, Olympia, WA 98501, unpubl. data). Locality numbers are given in Table 6.


Figure 9
Figure 9. Multidimensional scaling and minimum spanning tree (a tree connecting nearest genetic neighbors) of Rogers' genetics distances, based on 10 polymorphic loci (see text), between samples of even-year pink salmon from western Alaska, the Aleutian Islands, central and southeastern Alaska, British Columbia, and the Snohomish River, Washington. Locality numbers and sources of data as in Figure 8.

We then analyzed a set of data that combined the Washington sample and the samples from British Columbia to estimate the relationship of the Snohomish River sample to its nearest geographic neighbors to the north. However, genetic data for even-year spawners on southern Vancouver Island and in the Fraser River were not available; as noted in an earlier section, such fish are rare in these areas. Excluding the Alaska samples allowed us to include more loci in our analysis. We were aware that in British Columbia, Beacham et al. (1988) found highly significant allelic frequency differences for most polymorphic loci among even-year populations within three areas: 1) the south (localities 2-13 in Table 6) and central (localities 14-22) coasts of British Columbia, 2) the north (localities 23-27) coast of British Columbia, and 3) the Queen Charlotte Islands (localities 28-34). The amount of heterogeneity among areas was substantially greater than that among localities within areas.

In our analysis of unadjusted allelic frequencies, the UPGMA clustering of Rogers genetic distances (based on 15 loci: sAAT-3*, ADA-2*, CK-A1*, GPI-B1*, G3PDH-1*, mIDHP-1*, LDH-B1*, LDH-C*, sMDH-A1*, sMDH-B1*, mMEP-1*, PEPB-1*, PEPD-2*, PGDH*, and PGM-2*) placed the Snohomish River sample outside the cluster of British Columbia samples (Fig. 10). The Snohomish River sample also appeared as an outlier in the multidimensional scaling of these genetic distances (Fig. 11). In the analysis of adjusted allelic frequencies, the Snohomish River sample was embedded among samples from southern and central British Columbia in the UPGMA phenogram (Fig. 12). The Snohomish River sample showed a closer genetic affinity to samples from eastern Vancouver Island and central British Columbia when the multidimensional scaling was based on adjusted allelic frequencies (Fig. 13) than when the frequencies were unadjusted (Fig. 11). The analysis of Beacham et al. (1988), which used Nei s genetic distance, showed three distinct groups of even-year spawning pink salmon in British Columbia: 1) Queen Charlotte Islands, 2) north and central coasts, and 3) south coast and Vancouver Island. In our analysis using Rogers distance, groups 2 and 3 were less distinct.

If the Snohomish River population had experienced a strong reduction in population size at or since the time of its founding, a reduction in the amount of genetic variability might be apparent. Average heterozygosities with frequencies adjusted for sAAT-3* and PGDH* ranged from 0.134 to 0.171 among the British Columbia even-year pink salmon samples and was 0.134 in the Snohomish River sample. The Snohomish River population is at the lower end of this range of heterozygosities but does not appear to have lost a substantial amount of genetic variability relative to other populations. Therefore, if the Snohomish River even-year population experienced a reduction in population size or a founder effect, it apparently was not severe or protracted.

A hierarchical gene diversity analysis of 15 polymorphic loci (including adjusted frequencies for sAAT-3* and PGDH*) indicated that 98.5% of the total genetic diversity in British Columbia and Washington samples of even-year pink salmon was contained on average within populations, 0.9% was due to allelic frequency variability among populations within the areas, and 0.6% was due to differences among the 5 areas. This analysis suggests that the lossof a single population would not lead to substantial loss of overall genetic diversity as detected by allozymes in even-year pink salmon. The partitioning of genetic variability at other parts of the genome is less well understood, not as easily quantified, and may be different.


Figure 10
Figure 10. UPGMA phenogram of Rogers' genetic distances, based on 15 polymorphic loci (see text), and unadjusted allelic frequencies for sAAT-3* and PGDH (see text) between samples of even-year pink salmon from British Columbia (Beacham et al. 1988), and the Snohomish River, Washington (J. Shaklee, Washington Department of Fish and Wildlife, P.O. Box 43151, Olympia, WA 98501, unpubl. data). Locality numbers are given in Table 6.


Figure 11
Figure 11. Multidimensional scaling and minimum spanning tree of Rogers' genetics distances, based on 15 polymorphic loci (see text), and unadjusted allelic frequencies for sAAT-3* and PGDH between samples of even-year pink salmon from British Columbia and the Snohomish River, Washington. Locality numbers and sources of data as in Figure 10


Figure 12
Figure 12. UPGMA phenogram of Rogers' genetic distances, based on 15 polymorphic loci (see text), and adjusted allelic frequencies for sAAT-3* and PGDH (see text) between samples of even-year pink salmon from British Columbia and the Snohomish River, Washington. Locality numbers are given in Figure 10.


Figure 13
Figure 13. Multidimensional scaling and minimum spanning tree of Rogers' genetics distances, based on 15 polymorphic loci (see text), and adjusted allelic frequencies for sAAT-3* and PGDH between samples of even-year pink salmon from British Columbia and the Snohomish River, Washington. Locality numbers and sources of data as in Figure 10

Odd-year Pink Salmon: Genetic Variability Among Regions

Odd-year spawners occur throughout the range of pink salmon, but they are more abundant in the southern parts of this range in both Asia and North America. Several researchers have examined the genetic structure of odd-year populations in various regions, and some of these studies have included samples from other geographic regions; thus, a preliminary comparison of within- and among-region variability is possible. Shaklee and Varnavskaya (1994) examined 8 samples of Russian odd-year spawners and compared them with 15 samples of odd-year and 1 sample of even-year pink salmon collected from localities extending from southeastern Alaska, through British Columbia to Washington. Their cluster analyses of genetic distances based on 33 variable loci revealed groups of populations from three geographic areas: 1) Russia, 2) northern North America, including southeastern Alaska and northern British Columbia, and 3) southern British Columbia and Washington. The Russian and northern North American groups were genetically more closely related to one another than either was to the southern North American group. These analyses indicated the existence of a genetic discontinuity between northern and southern odd-year British Columbia populations that is larger than the discontinuity between Alaskan and Russian populations.

Similar results were found by Varnavskaya and Beacham (1992), who combined their allelic frequency data with those of Johnson (1979), McGregor (1983), and Beacham et al. (1988) to calculate genetic distances between samples based on five loci in common to these studies. Varnavskaya and Beacham s cluster analysis of genetic distances indicated that the odd-year spawners of Kodiak Island, southeastern Alaska and northern British Columbia were more closely related to odd-year spawners of the Kamchatka Peninsula than they were to odd- year spawners in southern British Columbia. However, these authors found that genetic distances between these groups were small relative to the distances between odd- and even-year spawners.

Odd-year Pink Salmon: Genetic Variability Within Regions

Asia--Glubokovskii and Zhivotovskii (1986) proposed a fluctuating stock model for the population structure of Russian pink salmon in an attempt to explain variability in genetic differentiation among these populations. They advanced this model as an explanation for periodic changes in population structure that are due to fluctuations of intensity of [genetic] exchange...between populations which is caused by the appearance of new and the disappearance of old migrational barriers (both natural and anthropogenic). The model predicts that genetic differences (as measured with allozymes) among these populations are unstable because of shifts in impediments to gene flow. These authors suggested that population structure in pink salmon follows this model for two reasons: 1) the abundances of fish in the same broodline changed markedly over time in an area, especially in the Kurile and Sakhalin Islands and on the Kamchatka Peninsula (Ivankov 1986); and 2) early genetic studies (Altukhov et al. 1983, Utter et al. 1980) indicated low levels of genetic differentiation among populations. Early genetic studies of Asian pink salmon (Kartavtsev et al. 1981, Salmenkova et al. 1981) failed to find the degree of genetic subdivision among populations that was present in other species of salmon. However, more recent studies of Asian populations (see below), have revealed a degree of genetic subdivision among populations that is similar to that observed among populations in North America. The agreement of the earlier results with the expectations of the fluctuating stock model may be largely due to analyses involving few loci, because the more comprehensive recent analyses are at odds with these expectations.

Four recent studies examined Asian populations in more detail. First, Kartavtsev (1991) studied allozymic variability at 5 loci in samples from 22 Asian rivers over 3 to 5 generations and found little allelic frequency heterogeneity among samples within 4 areas: 1) western Sakhalin Island, 2) eastern Sakhalin Island, 3) the Sea of Okhotsk, and 4) the Kamchatka Peninsula. Kartavtsev found significant differences between some of these areas in some years but not others, and concluded that extensive gene flow between localities was responsible for the genetic homogeneity among populations.

Second, Kartavtsev et al. (1992) extended the analysis of these data and found no significant deviations from expected Hardy-Weinberg genotypic proportions in samples pooled over geographic areas. Although the fit to Hardy-Weinberg proportions in the pooled sample was consistent with a lack of geographic differentiation among populations, this approach for detecting genetic differences among samples is weak.

Third, Varnavskaya and Beacham (1992) studied allozyme variability at 12 loci in pink salmon from 8 rivers on the east coast of the Kamchatka Peninsula. These samples were collected at river mouths, however, and therefore may not represent spawning-ground populations. G-tests detected significant (P < 0.05) overall allelic frequency heterogeneity among samples at five loci, and randomization tests detected significant heterogeneity at three loci. Cluster analysis of genetic distances showed that the samples from the Hailula and Uka Rivers were distinct from the other samples, but the genetic distances between these and the remaining samples were small.

Finally, in a larger study, Shaklee and Varnavskaya (1994) examined the products of 44 protein-encoding loci and found variability at 24 loci in 8 samples of odd-year spawners collected in 1991 from 8 localities in the Sea of Okhotsk, the Kamchatka Peninsula, and the western Bering Sea. None of these localities had been represented in the work of Varnavskaya and Beacham (1992). A geographically nested contingency-table analysis of allelic frequencies at 23 loci demonstrated significant total heterogeneity among the 8 samples for 3 loci, but the sum of the G-test statistics over variable loci was not significant. Shaklee and Varnavskaya (1994) detected little heterogeneity within the three regions and found only a single locus that showed significant differences among the three regions. Multidimensional scaling of Cavalli- Sforza and Edwards (1967) chord genetic distances between samples failed to show any geographically meaningful relationships among samples.

British Columbia--Beacham et al. (1985) collected spawning-ground samples from 4 rivers in British Columbia in 1982 (even-year spawners) and from 21 rivers in British Columbia and Washington in 1983 (odd-year spawners). Samples of odd-year spawners were divided into three geographic areas for analysis: 1) Johnstone Strait and Strait of Georgia (nine localities), 2) Fraser River (seven localities), and 3) Puget Sound (four localities). Significant allelic frequency heterogeneity was detected within each area, but the sums of G-test statistics over loci indicated that the greatest amount of heterogeneity was due to differences among localities in Johnstone Strait and the Strait of Georgia. Cluster analysis of genetic distances based on 14 polymorphic loci detected 3 groups corresponding to the groups used in the contingency-table tests, except that the sample from the south coast Indian River, just north of the Fraser River, was included with the Fraser River samples and not with the Johnstone Strait samples. The Fraser River and Puget Sound samples were more closely related to one another than either was to the samples from northern British Columbia.

Beacham et al. (1988) examined a much larger number of samples at 33 even-year and 47 odd-year spawning sites in British Columbia and 4 odd-year sites in Puget Sound. To test for allelic frequency differences, they divided odd-year spawners into 5 areas and found a significant degree of allelic frequency heterogeneity among localities for most of the 15 polymorphic loci examined. As was the case for even-year pink salmon (see above), on a larger geographic scale they found substantially greater heterogeneity in odd-year pink salmon populations among areas than among localities within areas. Cluster analysis of genetic distances between the British Columbia samples (not all of the loci were examined in Puget Sound samples) showed two somewhat distinct Canadian groups: a cluster including most of the northern island and mainland localities, and a cluster including Fraser River, Vancouver Island, and south coast localities. However, the cluster of northern samples was heterogeneous as it also contained five samples from southern British Columbia.

Washington--Shaklee et al. (1991) examined allelic frequency variability for 21 variable loci in 26 odd-year spawning localities in British Columbia and Washington in 1985, 1987, and 1989. A total of 52 collections were grouped into 26 samples by pooling multiple- year data at several localities, since little temporal variability was present among samples taken from different generations at the same locality. Samples from minor tributaries were also pooled in Hood Canal, the Snohomish River, the Stillaguamish River, the Skagit River, and the Nooksack River, since little variability was present among tributaries within these river systems. An analysis of chord genetic distances between samples revealed three geographic clusters: 1) north coast British Columbia and northern populations from south coast British Columbia; 2) southern populations from south coast British Columbia and Fraser River and Puget Sound (except Nooksack River); and 3) Hood Canal and Washington Strait of Juan de Fuca and Nooksack River (Fig. 1 in Shaklee et al. 1991).

The genetic definitions of groups 1 and 2 by Shaklee et al. (1991) were consistent with those of Beacham et al. (1985, 1988) to the extent that localities in the two studies overlapped. The second group consisted of three geographic subgroups: a) south coast of British Columbia on the Strait of Georgia, b) Fraser River and its tributaries, and c) Puget Sound. The samples from Hood Canal/Strait of Juan de Fuca (Olympic Peninsula) had not been examined by Beacham et al. (1985, 1988), but in the analysis by Shaklee et al. (1991) they appeared to represent a distinguishable cluster lying outside the Puget Sound, Fraser River and southern British Columbia clusters. The sample from the Nooksack River (northern Puget Sound) clustered with those from Hood Canal. Shaklee et al. (1991) hypothesized that the genetic similarity between the Nooksack River and Hood Canal populations reflected a supplementation of the natural Nooksack River population with eggs from the Hood Canal Hatchery in 1977. However, a subsequent analysis suggested that the Nooksack River population may be naturally distinct from other odd-year populations in Washington and southern British Columbia (J. Shaklee - footnote 25, Shaklee et al. 1995).

Shaklee et al. (1995) also reported the results of analyses of samples from the upper and lower Dungeness and Nisqually Rivers, so that allelic frequencies are now available for 19 naturally spawning populations in Washington. No genetic data were available for pink salmon from the Elwha River, where they may be extinct, or from the South Fork of the Nooksack River, where odd-year spawners are known to occur. We conducted our own analyses of Washington samples by combining them with British Columbia samples processed in the same laboratory (Shaklee et al. 1991). Allelic identities among these samples are therefore consistent. Since some river systems were represented by a single sample, samples of Washington fish were not pooled by river system; this configuration ensured that the full range of interpopulation variability would appear in the analyses.

Considering the 19 Washington sampling localities alone, we calculated 4 different genetic distances based on 23 polymorphic loci (sAAT-3*, sAAT-4*, ADA-2*, mAH-4*, sAH*, ALAT*, CK-A1*, CK-C1*, GPI-B2*, GPI-A*, G3PDH-1*, FDHG*, GDA*, sIDHP-2*, LDH-A1*, LDH-B1*, sMDH-A1,2*, sMDH-B1,2*, MPI*, PEPD-2*, PEP-LT*, PGDH*, and PGM-2*), and used UPGMA cluster analysis and multidimensional scaling analysis to examine relationships among samples. The results distinguished three Washington groups: 1) Dungeness River, 2) Hood Canal, and 3) Puget Sound (Fig. 14, UPGMA). In the UPGMA tree of chord genetic distances, the two samples from the Nooksack River clustered with Hood Canal samples, as found by Shaklee et al. (1991), and the sample from the Nisqually River was positioned outside the Hood Canal and Puget Sound clusters. Multidimensional scaling of chord genetic distances (Fig. 15) showed three groups, but multidimensional scaling of other genetic distances was less successful in distinguishing Hood Canal and Puget Sound samples as separate clusters.

The minimum-spanning tree, in which branches connect nearest genetic neighbors, indicated that samples from both the Nooksack and Nisqually Rivers are probably outliers from the Puget Sound group. None of the populations showed greatly reduced levels of genetic variability relative to the other populations that would indicate strong recent reductionsin population size. A hierarchical gene diversity analysis of allelic frequencies for the 19 Washington populations indicated that geographic subdivisions accounted for a very small proportion of the total diversity: 98.7% of the total allelic frequency variability was contained on average within populations; 0.3% was due to differences among tributaries within rivers; 0.5% was due to differences among rivers within regions; and 0.5% was due to differences among regions.


Figure 14
Figure 14. UPGMA phenogram of Cavalli-Sforza and Edwards' (1967) chord genetic distances, based on 23 polymorphic loci (see text) between samples of odd-year pink salmon from Washington (Shaklee et al. 1991; J. Shaklee, Washington Department of Fish and Wildlife, P.O. Box 43151, Olympia, WA 98501, unpubl. data). Localities (abbreviation): Lower Dungeness (LD), Upper Dungeness (UD), Grey Wolf (GW), Dosewallips (Do), Duckabush (Du), Hamma Hamma (ha), Middle Fork Nooksack (MN), North Fork Nooksack (NN), Nisqually (Ni), Puyallup (Pu), Sauk (Sa: Skagit), Skykomish (Sy: Snohomish), Bacon (Bc: Skagit), Snohomish (Sh), South Fork Stillaguamish (SSt), Skagit (Sk: main stem), Finney (Fi: Skagit), North Fork Stillaguamish (NSt), and Snoqualmie (Sq: Snohomish)


Figure 15
Figure 15. Multidimensional scaling and minimum spanning tree of chord genetic distances, based on 23 polymorphic loci (see text) between samples of odd-year pink salmon from Washington. Localities and sources of data as in Figure 14.


To put genetic variability among Washington odd-year spawners into a broader geographic perspective, we conducted additional analyses, adding 17 samples from British Columbia (Shaklee et al. 1991). This combined set of allelic frequencies for 20 polymorphic loci (sAAT-3*, sAAT-4*, ADA-2*, mAH-4*, sAH*, ALAT*, CK-A1*, CK-C1*, GDA*, GPDH- 1*, FDHG*, sIDHP-2*, LDH-A1*, sMDH-B1,2*, mMEP-1*, MPI*, PEPD-2*, PEP-LT*, PGDH*, and PGM-2*) included 36 populations extending from northern British Columbia to Puget Sound. A UPGMA cluster analysis of chord genetic distances between populations (Fig. 16) again indicated a major subdivision between north coast British Columbia samples and all other southern samples--the same result reported by Beacham et al. (1988) and Shaklee et al. (1991). Four clusters appeared among the southern samples: 1) Olympic Peninsula, 2) south coast British Columbia, 3) Fraser River, and 4) Puget Sound. One difference between this tree and the preceding one was in the positions of the Hood Canal and Nooksack River samples, both of which clustered more closely with the Dungeness River samples in the larger set of data (Fig. 16). Multidimensional scaling of the chord genetic distances (Fig. 17) revealed the major discontinuity between southern and northern British Columbia populations, but southern populations were distributed more or less continuously with small discontinuities between 1) Dungeness River, 2) Hood Canal, 3) Puget Sound, 4) Fraser River, and 5) south and central coasts of British Columbia. In this analysis, the Nooksack River sample was manifest as an outlier from the Hood Canal cluster rather than from the Puget Sound cluster, as it was in the preceding analysis of Washington samples alone. This change in configuration resulted from the use of overlapping but different samples of loci in the two analyses, and indicates that the genetic relationship of the Nooksack River population to other Washington and British Columbia populations is not well resolved.

The results of the gene diversity analysis did not change substantially with the addition of the Canadian samples in the analysis. Most (97.9%) of the variability was contained, on average, within populations; 0.2% was due to variability among localities within rivers; 0.8% was due to differences among rivers within regions; 0.7% was due to differences among regions; and 0.3% was due to differences between north coast British Columbia samples and the other southern samples.

The island model of migration can be used to estimate the number of migrants between each population in this analysis by assuming that allelic frequencies at all loci reflect an equilibrium between the differentiating effect of random genetic drift and the homogenizing effect of gene flow (Wright 1951). The FST value (a measure of genetic differentiation among populations), averaged over all loci, was 0.021 and yielded an estimate of 11-12 migrants into each population per generation. This measure provides a rough estimate of the level of intermixing between populations. However, if the populations are not at equilibrium (i.e., if insufficient time has elapsed since population expansion during the current interglacial period), then this statistic overestimates gene flow between populations. On the other hand, this approach may underestimate the number of migrants into salmon populations because salmon straying does not strictly follow the island model of migration.


Figure 16
Figure 16. UPGMA phenogram of chord genetic distances, based on 20 polymorphic loci (see text) between samples of odd-year pink salmon in Washington and British Columbia. Localities and sources of data as in Figure 14, with the following additional localities from British Columbia (abbreviation): Fraser River (FR: main stem), Vedder (Ve), Harrison (Hr), Coquihalla (Co), Thompson (Th), Seton (Se), Bridge (Br), Indian (In), Skwawka (Sw), Quinsam (Qu), Kakweiken (Ka), Phillips (Ph), Wakeman (Wa), Adam (Ad), Keogh (Ke), Babine (Ba) and Andesite (An).


Figure 17
Figure 17. Multidimensional scaling and minimum spanning tree of chord genetic distances, based on 20 polymorphic loci (see text) between samples of odd-year pink salmon Washington and British Columbia. Localities and sources of data as in Figure 16.

Conclusions

Genetic data do not show a close affinity between the single population of even-year pink salmon in the Snohomish River in Washington and any known source of eggs translocated to Washington from northern British Columbia and Alaska. The analysis of available data indicates that the Snohomish River even-year pink salmon population is genetically most closely related to central and south coast British Columbia populations and is a peripheral population at the southern extreme of the geographic distribution of even-year spawners. However, genetic data are not available for even-year spawners from southern Vancouver Island and the Fraser River, and the Snohomish River population may be closely related to these populations.

Odd-year pink salmon populations located around the Pacific Rim can be divided into three major genetic groups: 1) Asia, 2) Alaska and northern British Columbia, and 3) southern British Columbia and Washington. Populations in the first two groups are more closely related to one another than they are to populations in the third group. These major divisions may have resulted from the formation of glacial barriers during the last ice age (Beacham et al. 1988). The populations within each region are also genetically subdivided to some degree, and the degrees of subdivision among populations within each of the three regions are about the same.

Four subgroups of odd-year spawners are distinguishable within southern British Columbia and Washington: 1) Olympic Peninsula, 2) south coast of British Columbia, 3) Puget Sound, and 4) Fraser River. Groups 3 and 4 are most closely related to each other, and groups 2 and 1 are progressively more distantly related to groups 3 and 4. Within Washington, the populations of the Nooksack and Nisqually Rivers are genetic outliers and do not fall within expected geographic clusters in the analyses. The relationship of the Nooksack River population to other populations depended on the sample of loci and populations used in the analyses: in analyses limited to samples from Washington, it was genetically closest to Puget Sound populations, but when analyses included British Columbia samples and a different set of loci, it tended to show greater affinity to Hood Canal populations. Consequently, the relationship of this population to other populations in the region is unclear.

It should be recognized that the conclusions drawn from these analyses must be tempered by the fact that the analyses were generated with data from several studies with different sets of loci and different means of generating and interpreting electrophoretic data. Some of the analyses involved an implicit assumption that the different sets of data are compatible, an assumption that has been shown to be invalid for some pink salmon studies (White and Shaklee 1991).

Discussion and Conclusions on ESU Determinations

Based on the genetic, life history, and ecological information presented above, the Biological Review Team identified two ESUs for North American pink salmon in Washington and southern British Columbia. In the following discussion we describe these ESUs and outline the issues that were valuable to the BRT in making each ESU determination.

Even-year Pink Salmon

A single population of even-year pink salmon occurs in the United States south of Alaska, in the Snohomish River. This population is genetically much more similar to even-year pink salmon from British Columbia and Alaska than it is to odd-year pink salmon from Washington. This pattern of similarity is also found in life history traits such as body size and run timing. This result is consistent with numerous other studies that have found large genetic differences between even- and odd-year pink salmon from the same area, with the magnitude of the differences roughly comparable to that found between coastal and inland steelhead, Oncorhynchus mykiss (Okazaki 1984, Reisenbichler et al. 1992). The BRT concluded, therefore, that Snohomish River even-year pink salmon are in a different ESU than odd-year pink salmon from Washington.

The origin of the Snohomish River even-year population and its relationship to other even-year populations in British Columbia and Alaska is less certain. Although several concerted efforts were made between 1910 and 1956 to introduce even-year pink salmon from Alaska and British Columbia into Puget Sound, no direct evidence indicates that any were successful in producing sustained returns. The allozyme studies described here also failed to show a similarity of the Snohomish River sample to any of the putative sources of stock transfers for which data are available.

Although available data do not provide evidence to support the hypothesis that the Snohomish River population has resulted from a human-mediated stock transfer in this century, this possibility cannot be excluded entirely. Not all sources of even-year stock transfers are known, and some of those that are known have not been characterized genetically. The genetic distinctiveness of the current Snohomish River population might have resulted from genetic change (founder effect or subsequent genetic drift or both) associated with a natural colonization event, but it might also have occurred relatively recently as a result of similar processes in an introduced population. In addition, available genetic information for the Snohomish River sample is based on a single year s collection.

Nevertheless, the BRT concluded that even-year pink salmon in the Snohomish River should be presumed to be a native population in the absence of more convincing evidence that it is not. This decision was reached because 1) observed genetic and life history traits of this population are consistent with what might be expected from a natural colonization, and 2) this population has apparently been naturally self-sustaining in the Snohomish River for at least 8 generations (and possibly for 25 or more).

At present, the Snohomish River population is relatively small (up to a few thousand adults per generation), and this raises the issue of the importance of historical population size in considering and defining ESUs. Because ESUs are intended to represent units that are largely independent from other such units over evolutionarily important time frames, Waples (1991a, p. 19) argued in the NMFS Definition of Species paper that

A Pacific salmon population should not be considered an ESU if the historic size (or historic carrying capacity) is too small for it to be plausible to assume the population has remained isolated over an evolutionarily important time period. In making this evaluation, the possibility should be considered that small populations observed at present are still in existence precisely because they have evolved mechanisms for persisting at low abundance.
The small size of the current Snohomish River even-year population suggests that it may be part of a larger geographic unit on evolutionary time scales (hundreds or thousands of years). However, the Snohomish River odd-year population, which has the same spawning habitat available, is 1-2 orders of magnitude larger, so it is possible that the even-year population was also larger in the past. If so, long-term persistence of the even-year population in isolation from other even-year populations would be easier to explain.

The Snohomish River even-year pink salmon population is geographically isolated by several hundred kilometers from other even-year populations of appreciable size. However, life history features of the Snohomish River even-year population are similar to those in other even-year populations from central British Columbia. For example, peak spawning time of Snohomish River even-year pink salmon is comparable to that of some British Columbia even- year pink salmon, yet it is about 2 weeks earlier than that of Snohomish River odd-year pink salmon. This timing differs despite spatial overlap in habitat use by even- and odd-year adults in that system. Results of genetic analyses depend heavily on whether an adjustment is made for possible differences between laboratories in methods for recording the data. A comparison of published data suggests that the Snohomish River even-year population is genetically the most distinctive of any sample from the United States or southern British Columbia. However, the Snohomish River population is much less distinctive in analyses of adjusted allelic frequencies.

The BRT could not resolve with any degree of certainty the extent of the ESU that contains the Snohomish River even-year pink salmon population. This issue was particularly difficult because it is not clear which analyses--those with or without the adjustment for possible bias--should be preferred. After considering all available information, about half the BRT members concluded that the Snohomish River even-year population belongs in an ESU by itself, and half concluded that it belongs in an ESU with populations from British Columbia. Most of those favoring the latter scenario felt that the ESU probably includes all even-year pink salmon from streams entering the Strait of Georgia and Johnstone Strait.

In any case, the BRT was unanimous in agreeing that any conclusion about the extent of the even-year pink salmon ESU should be regarded as provisional and subject to revision should substantial new information become available on 1) the Snohomish River population, 2) the history and success of transplanting even-year pink salmon into northwestern Washington, or 3) small populations of even-year pink salmon in southern British Columbia that are at present poorly understood.

Odd-year Pink Salmon

Genetic, life history, and environmental data were the most important factors in consideration of ESUs for odd-year pink salmon. Environmental and ecological characteristics generally show a strong north-south trend, but no substantial differences were identified that consistently differentiated Washington and British Columbia populations. An east-west gradient, separating populations along the Strait of Juan de Fuca from those to the east, was considered more important for evaluating pink salmon populations. In addition, three U.S. rivers supporting odd-year pink salmon populations (the Nisqually, Nooksack, and Puyallup Rivers) are dominated by glacial runoff, which may promote special adaptations in spawning populations. However, glacial rivers are also common in most other areas of pink salmon distribution throughout the Pacific Rim.

Although odd-year pink salmon show considerable variation in body size among populations in Washington, the range of this variation does not exceed that found in British Columbia. Among U.S. populations, Nooksack River fish average the smallest and Hood Canal fish the largest in body size. Time of peak spawning varies over about a 4- to 5-week period in northwestern Washington; spawning occurs earliest in the Nooksack and Upper Dungeness Rivers, followed by Hood Canal and then Puget Sound (excluding the Nooksack and Nisqually Rivers), with the latest spawning occurring in the Nisqually River.

Some pink salmon appear to reside in Puget Sound during their marine phase rather than migrating to the open ocean (Jensen 1956, Hartt and Dell 1986). This behavior has not been documented for pink salmon in other areas, but this may simply reflect a lack of appropriate studies. Furthermore, resident Puget Sound fish have not yet been associated with any particular freshwater population(s), so the importance of this trait in helping to define odd-year pink salmon ESUs is not clear.

Comprehensive genetic analyses show that odd-year pink salmon from southern British Columbia and Washington are clearly in a different evolutionary lineage than nearby even-year populations and more northerly odd-year populations. Within the southern British Columbia- Washington group, there is also evidence of geographic population genetic structure, with some differences among populations from the Dungeness River, Hood Canal, Puget Sound, the Fraser River, and southern and central British Columbia. In some analyses, samples from the Nisqually and Nooksack Rivers were both genetic outliers but were not similar to each other. However, none of the genetic differences found within the southern British Columbia- Washington group are very large in absolute magnitude. For example, the FST value (averaged over all loci) for all odd-year samples of pink salmon from Puget Sound to central British Columbia was only 0.021, considerably smaller than the value found among populations within the ESU for Snake River spring/summer chinook salmon (O. tshawytscha) (FST = 0.034; Waples et al. 1993) or among populations of steelhead (O. mykiss) within the Klamath Mountains Province ESU (FST = 0.033; NMFS, unpubl. data - footnote 26).

Although genetic differences (as determined by protein electrophoresis) among odd-year pink salmon populations in the Puget Sound/British Columbia area are relatively modest, the general geographic coherence of the structure that does exist indicates that there has been some reproductive isolation among pink salmon populations within this larger area. The BRT next considered whether any of these individual groups of populations (Dungeness River, Hood Canal, Puget Sound, Fraser River, southern and central British Columbia) might represent a substantial contribution to ecological/genetic diversity of the species as a whole. As noted above, the BRT did not find strong environmental or ecological differences among these areas. Two of the U.S. rivers dominated by glacial runoff (Nooksack and Nisqually) also have pink salmon populations that are genetically somewhat distinctive. However, the two populations are not similar to each other, either on the basis of genetic or life history characteristics, so it seems unlikely that they are evolving together as a unit independent from other odd-year pink salmon populations. A minority of the BRT believed that these two populations were distinctive enough to each be separate ESUs, but most members thought that they simply represented part of the natural variability within a larger ESU.

There was somewhat stronger concurrence among BRT members that odd-year pink salmon populations from the Strait of Juan de Fuca (i.e., those in the Dungeness and Elwha Rivers) were in a separate ESU from other Puget Sound and British Columbia odd-year populations. Evidence to support this view is based on genetic differences detected by protein electrophoresis, geography and habitat differences, and the fact that recent declines (including the possible extinction of the Elwha River population) are shared by Strait of Juan de Fuca populations but not Puget Sound populations. Notably, the upper Dungeness River population shows an extraordinarily early run timing for the region and an apparently unique maturation strategy. However, no life history data are available for the Elwha River population, and the lower Dungeness River population has life history features similar to those of other Puget Sound populations. Furthermore, although there are some habitat and ecological differences between the Dungeness and Elwha Rivers and others in the Puget Sound area, a stronger environmental transition is found west of the Elwha River, where much wetter areas typical of the Olympic Peninsula are encountered (Weitkamp et al. 1995). The majority of the BRT therefore concluded that the populations of the Dungeness and Elwha Rivers (if the latter still exists) are part of the larger Puget Sound/British Columbia ESU for odd-year pink salmon.

Based on currently available information, the BRT concluded that the northern boundary of the Puget Sound/southern British Columbia ESU corresponds to the geographic location where a strong genetic discontinuity was observed in the Johnstone Strait region of British Columbia. The ESU does not include northern British Columbia, Alaskan, or Asian populations. In Washington, the westernmost populations in this ESU are both found in the Dungeness River, but the ESU presumably would also include the Elwha River population, if a remnant still exists. The BRT felt there was insufficient information to determine whether other populations on the Olympic Peninsula or farther south (if any such populations exist) would be included in this ESU. Also uncertain is the relationship of odd-year populations from southwestern Vancouver Island to this ESU.

It is clear that pink salmon populations within the Puget Sound/southern British Columbia ESU contain a substantial amount of genetic and life history diversity. Much of this diversity is contained within the marginal populations in Washington, which represent the southernmost regularly spawning populations of North American pink salmon and are likely to be important for colonizing available habitat in this region when conditions are favorable to do so (Scudder 1989, Lesica and Allendorf 1995). The ESU has representatives that are distinctive on the basis of run timing and maturation strategy, and may include an unknown population or populations that lack an oceanic migration, using instead a protected embayment for the marine phase of development and growth. In addition, the ESU has populations that spawn far from tidewater, in stream reaches heavily influenced by glacial runoff.


Table of Contents