Previous Studies of Population Genetic Structure
Protein electrophoresis--Allendorf (1975) first distinguished two major groups of O. mykiss in Washington, Oregon, and Idaho, separated geographically by the Cascade Crest; he termed these groups inland and coastal. These two groups have large and consistent differences in allele frequency that apply to both anadromous and resident forms. Subsequent studies have supported this finding (Utter and Allendorf 1977, Okazaki 1984, Schreck et al. 1986, Reisenbichler et al. 1992), and similar differences have been identified between O. mykiss from the interior and coastal regions of British Columbia (Huzyk and Tsuyuki 1974, Parkinson 1984).
Several genetic studies since the mid-1970s have used protein electrophoresis to examine population structure in coastal or inland O. mykiss. Allozyme studies of coastal Oregon steelhead have been reported by Hatch (1990) and Reisenbichler et al. (1992). Hatch (1990) surveyed 13 protein-coding loci in steelhead from 12 hatcheries and 26 coastal rivers or tributaries in Oregon. He found evidence for a north-south cline in allele frequencies in 5 of the 13 enzyme systems analyzed, but only in river basins larger than 350 km2. Hatch also reported that "the area south of the Coos River was marked by sharp transition in four enzymes..." (p. 17) and that "the pattern of several alleles ending their detectable Oregon presence just north of Cape Blanco suggests that there is a less than average amount of straying between the populations north and south of this feature" (p. 33).
Reisenbichler et al. (1992) examined 10 polymorphic gene loci in steelhead from 37 natural and hatchery populations in the Pacific Northwest, including 24 from the Oregon coast and two in northern California (Trinity River summer-run and Mad River Hatchery winter-run). They did not discuss clines in allele frequencies; instead, they found evidence for genetic differentiation between some clusters of populations. For example, steelhead north of the Umpqua River formed a separate cluster from steelhead in southern Oregon. The Trinity River sample was genetically similar to most of the Rogue River samples, but steelhead from the Mad River Hatchery were genetically distinct from other hatchery and natural populations in California and Oregon.
As part of previous ESA status reviews, NMFS biologists analyzed genetic variability at 39 polymorphic gene loci in 20 samples of coastal steelhead from the Nehalem River in northern Oregon to the Eel River in northern California (Busby et al. 1993, 1994). These studies found evidence for three genetic groups of populations in the area sampled: Oregon coast north of Cape Blanco (3 samples), Cape Blanco to the Klamath River Basin, inclusive (13 samples), and south of the Klamath River Basin (4 samples). Little geographic pattern was evident for samples from the area between Cape Blanco and the Klamath River. Redwood Creek, the first major stream south of the Klamath River, appears to be in a transitional zone; the sample from this stream is similar to the southern group but also has some genetic affinity with samples from the Klamath River and areas to the north. The sharp transition in allele frequencies for steelhead populations in this area is apparent in Figure 4 of Busby et al. (1994).
Berg and Gall (1988) examined genetic variability at 24 polymorphic loci in 31 California populations "known to have been inhabited by anadromous rainbow trout prior to the major water projects of the twentieth century" (p. 123). Twenty-three of their samples were from the upper Sacramento River Basin, three were from the upper Klamath River Basin, and the remaining five were from coastal streams from Eel River to San Diego County. Sample sizes averaged about 30 fish per population and ranged from a high of 57 to a low of 7. Berg and Gall (1988) found relatively high levels of genetic variability but no clear geographic patterns in the genetic relationships among populations.
Reisenbichler and Phelps (1989) found variation at 19 gene loci in steelhead from 9 drainages in northwestern Washington (primarily the Olympic Peninsula). However, they found genetic differences between drainages to be much smaller than had been reported by Parkinson (1984) for steelhead populations from adjacent drainages in British Columbia. Reisenbichler and Phelps (1989) and Reisenbichler et al. (1992) suggested that since both Washington and Oregon had far more extensive hatchery steelhead programs in the 1970s and early 1980s than did British Columbia, the relative homogeneity among populations in these states may be due to introgression of hatchery fish into naturally spawning populations. Furthermore, during that period, hatcheries in both Oregon and Washington predominately used steelhead that had originated from one or two within-state sources (the Alsea River stock in Oregon and the Chambers Creek and Skamania stocks in Washington). However, Hatch (1990) pointed out that the geographic area covered by the Reisenbichler and Phelps (1989) study (natural populations collected primarily from a 70 km stretch of coastline) might be too small to allow a direct comparison with the British Columbia study.
As part of a comprehensive effort to inventory wild stocks of anadromous salmonids, the Washington Department of Fish and Wildlife (WDFW) recently published a report of the first year of genetic analyses for steelhead populations. Phelps et al. (1994) reported new data for 56 variable gene loci for 12 natural and 8 hatchery populations, primarily in Puget Sound and the lower Columbia River. Furthermore, WDFW data for additional samples allowed the investigators to conduct analyses on 30 different populations. With few exceptions, usually involving hatchery stocks, Phelps et al. found statistically significant differences between all pairs of populations. This contrasts with results of Reisenbichler and Phelps (1989), who generally failed to find significant differences between populations on the Washington coast.
Phelps et al. (1994) used several different methods to examine population structure. One consistent result was a high degree of genetic similarity among samples from winter-run steelhead hatcheries, including those from Puget Sound (Skykomish River, Chambers Creek, Tokul Creek), Olympic Peninsula (Bogachiel River), and the Columbia River (Skamania, Beaver Creek). Relationships among the remaining Puget Sound samples were less clear. For example, the summer-run sample from Deer Creek showed affinities to winter-run fish from the North Fork Stillaguamish River (to which Deer Creek is a tributary), to summer-run fish from the Skykomish River Hatchery, or to no populations in particular, depending on the analysis.
Phelps et al. (1994) also considered data for 14 samples of steelhead from the Columbia River in their study. Summer-run samples from the Wind and Washougal Rivers in the lower Columbia River were outliers in the analyses. The Wind River sample contained an allele at a frequency of 15% that was not found in steelhead in any other sample analyzed by Phelps et al. (1994), and this presumably is responsible for the distinctiveness of the Wind River sample. As expected, Phelps et al. (1994) found that inland steelhead were genetically distinct from the samples of coastal steelhead examined. The inland group was represented primarily by six samples from the Klickitat River, with additional samples from Big White Salmon River, Satus Creek in the Yakima River Basin, and Wells Hatchery in the middle Columbia River. The relationships among these samples are difficult to determine from the results presented by Phelps et al. because the patterns of genetic affinity differed among the various analyses they used.
Phelps et al. (1994) examined their genetic data for evidence of the effects of hatchery fish on natural populations. The presence, in most cases, of statistically significant differences between the hatchery and natural samples of steelhead suggests that at least some native population structure remains. In addition, Phelps et al. found eight loci that had alleles at relatively uniform frequencies among the winter-run hatchery steelhead populations that could be used as indicators of the degree of introgression into natural populations. Based on this analysis, they concluded that the Cedar River, Deer Creek, North Fork Skykomish, North Fork Stillaguamish River, Wind River, Washougal River, and Big White Salmon River populations had limited amounts of hatchery introgression and that the Green River, Skykomish River main stem, Tolt River, Raging River, and Pilchuck River had moderate to large amounts of hatchery introgression. Because the "marker" alleles only occurred at frequencies of a few percent even in the hatchery steelhead stocks, these conclusions should be regarded as tentative.
Phelps et al. (1994) also found large genetic distances (about three times as large as the distance between inland and coastal steelhead) between four widely used rainbow trout hatchery stocks from Washington and all steelhead populations examined. They concluded that there has been little, if any, permanent genetic effect on the sampled steelhead populations from the widespread stocking of rainbow trout over the past century. Campton and Johnston (1985) found a different result for some O. mykiss populations in the Yakima River Basin, where they found evidence for introgression of non-native rainbow trout into wild populations. However, the affected populations were believed to be nonanadromous, and Campton and Johnston (1985) found no evidence for introgression of hatchery rainbow trout (or steelhead from Skamania Hatchery) into natural steelhead populations in the Yakima River.
Leider et al. (1995) reported preliminary results for an additional 55 samples of steelhead and wild resident rainbow trout from Washington. These samples considerably extended the geographic coverage in the WDFW data set for the Olympic Peninsula and southwest Washington coast. The most important result of the new samples is that they revealed considerably more geographic coherence to the population genetic structure of coastal steelhead in Washington than had been evident in previous studies. In the analyses of Leider et al. (1995), the patterns of genetic affinity among populations differed somewhat depending on the distance metric used, and some samples were outliers with no clear affinity to any group. In general, however, samples from the following geographic areas tended to be more similar to one another than they were to samples from other areas: north Puget Sound (including the Stillaguamish River and drainages to the north), south Puget Sound, Olympic Peninsula, southwest Washington, and lower Columbia River (Kalama, Wind, and Washougal Rivers). Notable genetic outliers included the Nooksack River and the Tahuya River. The genetic relationships among these geographic areas do not appear to be well resolved because the pattern of affinities differed substantially among analyses.
Inland O. mykiss were represented by 48 samples in the Leider et al. (1995) study. Analyses based on Nei's (1978) and Cavalli-Sforza and Edward's (1967) distances both found consistent differences between samples from the Yakima and Klickitat River Basins, and both analyses also showed that samples from Wells Hatchery were outliers within the inland group. No samples from natural populations in the upper Columbia River were included in the Leider et al. (1995) study. Leider et al. acknowledged some uncertainty in identifying the boundary between inland and coastal forms, but on the basis of genetic data tentatively placed it between the Wind and Big White Salmon Rivers.
Several other genetic studies have included steelhead from the Columbia River Basin. Reisenbichler et al. (1992) focussed on steelhead from coastal streams but also included 10 samples from the Columbia River Basin. Within their study, they found the greatest degree of genetic differentiation between the inland and coastal forms. Within the inland group, four samples from the Snake River and three from the Deschutes River formed separate genetic clusters. The three samples Reisenbichler et al. (1992) examined from the upper Willamette River formed the most distinctive subgroup within the coastal group.
The study of Schreck et al. (1986), which examined life history and morphological features as well as biochemical genetics, included the greatest number and geographic range of steelhead samples from the Columbia River of any study to date. Again, they found the largest differences between steelhead from east and west of the Cascades. Coastal forms from west of the Cascades could be further partitioned into a subgroup from the upper Willamette River, a subgroup from the lower Columbia River, and a subgroup containing samples from both the lower Columbia and Willamette Rivers. East of the Cascades, Schreck et al. also found evidence for differentiation among populations but only a weak geographic pattern to the observed structure.
Hershberger and Dole (1987) examined samples from nine populations of inland steelhead from tributaries of the Columbia River between Rock Island and Chief Joseph Dams. They found 20 polymorphic gene loci but relatively little genetic differentiation among populations from the Wenatchee, Methow, Entiat, and Okanogan Rivers. In contrast, they found relatively large allele frequency differences between these samples and a sample of coastal steelhead from the Skamania River.
Currens and Schreck (1993) examined genetic and meristic variation in adult steelhead used for broodstock in the Umatilla River and in samples from 13 populations of O. mykiss in the Umatilla River Basin. They found significant allele frequency differences among populations but no strong geographic patterns. Results suggested that steelhead from one population (McKay Creek) were the offspring of native and introduced rainbow trout. Currens and Schreck (1993) did not compare genetic data for the Umatilla River samples to data for other populations in the Columbia River Basin, but they did cite unpublished meristic data that distinguished Snake River steelhead from those in the middle and upper Columbia River.
Milner and Teel (1985) examined steelhead from 13 localities in the Snake River and found three major genetic clusters: one including four Salmon River samples, another including three samples from the Lochsa and Selway Rivers in the Clearwater River Basin, and a third including Dworshak National Fish Hatchery (NFH), several samples from the lower Clearwater River, and one sample each from the Grande Ronde and Imnaha River Basins.
Waples et al. (1993) summarized genetic data based on 50 polymorphic gene loci for 2 years of samples of steelhead from the Snake River. Results included the following: 1) The two samples from Dworshak NFH were the most distinctive genetically and have substantial allele frequency differences compared to all other natural and hatchery samples. 2) Natural samples from the Clearwater River differed somewhat from those from other drainages, and there was weaker evidence for differentiation between steelhead from the Grande Ronde, Imnaha, and Tucannon Rivers (Salmon River populations were not included in the experimental design). 3) In general, differences between temporal samples from the same stream were smaller than differences between geographic populations.
The steelhead population in Dworshak NFH is derived from native fish from the North Fork Clearwater River that were brought into the hatchery in 1969 when Dworshak Dam blocked access to their native habitat. To evaluate whether the distinctive genetic characteristics of Dworshak NFH steelhead might be the result of genetic changes in the hatchery (for example, spawn timing has been shifted in the hatchery, and there may have been selection for large fish, at least in the early years of the program), we examined genetic profiles for a limited number of gene loci scored in samples from the hatchery dating back to 1972 (Milner 1977 and Teel). These data show some variation over time but do not show a trend toward greater divergence from natural populations in more recent samples. Broodstock data (footnote 7) , which show that over 1,000 adults were spawned at the hatchery each year since 1969, also fail to provide evidence for a population bottleneck that might have caused substantial allele frequency changes due to drift. In the future, we hope to compare genetic profiles of Dworshak NFH fish with populations of resident O. mykiss in the North Fork Clearwater River, provided that native populations that have been largely unaffected by releases of hatchery rainbow trout into Dworshak Reservoir can be identified.
DNA--In recent years, genetic methods that analyze DNA variation directly have seen increasing use in salmonid studies, and we are aware of two studies of mitochondrial DNA (mtDNA) that assess population structure in steelhead. In a study that remains unpublished, Buroker examined restriction-fragment-length polymorphisms in mtDNA from 120 individuals from 23 major river systems from Alaska to California. He found no evidence for strong geographic structuring of populations, as most of the common clonal types were widely dispersed. However, Buroker also found that steelhead from southern Oregon were highly diverse in mtDNA. In the 120 fish analyzed, 18 different mtDNA clonal types were observed. These clones were clustered into four lineages, all of which overlap in southern Oregon. The 12 fish examined from the Rogue River had 6 of the 18 mtDNA clonal types observed in the study.
In another study, Nielsen (1994; see also Nielsen et al. 1994) sequenced part of the D-loop section of mtDNA in 37 samples of steelhead and rainbow trout in California and found that a different mtDNA clonal type was the most common in each of three geographic regions: north coast (Humboldt Bay to Gualala Point), central coast (Russian River to Point Sur), and south coast (San Simeon Point to Santa Monica Bay). These regions were defined through a combination of genetic and ecological (primarily ocean upwelling and plankton distribution) information (Nielsen) . Nielsen also found significant differences between the regions in allele frequencies at a nuclear DNA microsatellite locus.
Neeley (1995) performed some additional statistical analyses on Nielsen's data and some new mtDNA data collected specifically for the status review (Cramer et al. 1995). Neeley used principal components analysis to summarize variation at all 13 mtDNA alleles reported by Nielsen. The first two principal components together explained 70% of the total variation, and Neeley compared scores for each population on these two principal components to a ranked indicator of their latitude. Simple and multiple regression analyses suggested a partitioning of the populations into three groups based on latitude, with the boundary between the northern and central groups occurring just north of the Russian River, and the boundary for the central and southern groups occurring just south of the San Lorenzo River. Additional variation exists in the mtDNA data that is not explained by the first two principal components or the three-group partition, but no clear geographic patterns to this variation could be detected.
Chromosomal studies--Chromosomal karyotypes in steelhead and rainbow trout have also been extensively studied (see review in Thorgaard 1983). In a survey of steelhead from Alaska to central California, Thorgaard (1983) found that although chromosome numbers ranging from 58 to 64 were observed, a 58-chromosome karyotype was the most common in most samples. In contrast to results for studies of morphological and allozyme characters, Thorgaard did not find chromosomal differences between interior and coastal O. mykiss populations. All interior/redband trout populations had predominately 58 chromosomes, as did most coastal rainbow trout and steelhead populations.
The exceptions to the 58-chromosome pattern, however, provide insight into population genetic structuring in O. mykiss. Two geographic regions were characterized by steelhead with 59 or 60 chromosomes: the Puget Sound/Strait of Georgia region and the Rogue River/northern California region. However, the karyotypes of fish from these two regions were different; northern fish with 59 or 60 chromosomes had a different number of subtelocentric and acrocentric chromosomes than did southern fish (Thorgaard 1977). Farther south, winter steelhead in the Mad and Gualala Rivers from northern California and resident trout from the San Luis Rey River in southern California had 61-64 chromosomes (Thorgaard 1983).
Although Thorgaard's (1983) study showed that an unusual 60-chromosome karyotype exists in the Puget Sound region, sampling in that study was limited to a very few populations. Ostberg and Thorgaard (1994) examined additional populations in the area and found the 60- chromosome karyotype in presumed native steelhead from the Nooksack, Cedar, and Stillaguamish Rivers.
Comparison of Steelhead and Rainbow Trout
Allendorf (1975) found that the genetic distinction between coastal and inland O. mykiss applies to both life history forms; that is, rainbow trout east of the Cascades are genetically more similar to steelhead from east of the Cascades than they are to rainbow trout west of the Cascades. Many recent studies of O. mykiss have focussed on either rainbow trout or steelhead and thus provide no direct information about the relationship between the forms on a finer geographic scale. However, Leider et al. (1995) included several new samples of rainbow trout from the Elwha and Cedar Rivers in their study of steelhead populations in Washington and found that their results "support the hypothesis that the two forms were not reproductively isolated from each other." Leider et al. also concluded that, based on preliminary analysis of data collected previously for the Yakima and Big White Salmon Rivers (Pearsons et al. 1994, Phelps et al. 1990), wild resident rainbow trout in those streams would be indistinguishable from steelhead. In addition, some protein electrophoretic studies that have reported data only for rainbow trout probably also included samples of steelhead (K. Currens) . For example, in the John Day River, an Oregon tributary of the Columbia River, genetic differences between O. mykiss from the North and South Forks were larger than differences between presumed steelhead and rainbow trout in the South Fork (Currens et al. 1987). In the Deschutes River, another Oregon tributary of the Columbia River, Currens et al. (1990) found much larger genetic differences between O. mykiss from above and below a barrier falls, but relatively modest differences between presumed steelhead and rainbow trout from below the falls.
In a study of mtDNA in O. mykiss, Wilson et al. (1985) compared 19 steelhead from 4 locations in British Columbia with 19 rainbow trout from British Columbia, Alberta, and California. No genetic differences were detected between steelhead and rainbow trout from one British Columbia location (the lower Fraser River), but steelhead from the other three populations showed a greater genetic affinity to each other than to rainbow trout from any of the populations sampled. However, this result is difficult to interpret because of the small sample sizes and the fact that there were only two localities at which both steelhead and rainbow trout were collected. Furthermore, Buroker (footnote 8) found that the mtDNA marker Wilson et al. (1985) used to distinguish rainbow trout was the most common type found in his study of North American steelhead.
Gall et al. (1990) examined allozyme variation in resident O. mykiss from the San Leandro Creek watershed, which drains into the east side of the San Francisco Bay. These fish are believed to be descended from steelhead, which have not had access to this area since the construction of Chabot Reservoir in 1875. Gall et al. (1990) found that samples from two creeks upstream of the reservoir are genetically more similar to coastal O. mykiss than they are to inland forms or to hatchery rainbow trout. Nielsen (footnote 9) compared mtDNA haplotypes in southern steelhead with those in several California populations of resident O. mykiss and in several stocks of hatchery rainbow trout that have been stocked in coastal California streams. She found that some resident populations resemble nearby anadromous populations in their mtDNA profiles, but others show evidence of introgression from hatchery rainbow trout.
Differentiation based on timing of upstream migration in steelhead has also been investigated by genetic methods. Allendorf (1975) and Utter and Allendorf (1977) found that summer and winter steelhead of a particular coastal stream tended to resemble one another genetically more than they resembled populations of adjacent drainages with similar run timing. Later allozyme studies have supported these conclusions in a variety of geographical areas (Chilcote et al. 1980, Schreck et al. 1986, Reisenbichler and Phelps 1989), including the Rogue River (Reisenbichler et al. 1992). However, in each of these more recent studies, the summer-run stocks have had some extent of hatchery introgression and therefore may not represent the indigenous population. Furthermore, in at least some cases, interpretation of the results may be complicated by difficulties in determining run timing of the fish sampled.
Thorgaard (1983) analyzed chromosomal variability in winter- and summer-run steelhead from two rivers that had little history of hatchery introductions: the Quinault River in Washington and the Rogue River in Oregon. Chromosome number differed between the two river systems but was similar in summer and winter steelhead within each river system.
For this status review, two types of new studies were undertaken by NMFS to enhance our understanding of population genetic structure in west coast steelhead. First, new samples from Idaho and California were collected for allozyme analysis. Second, recent data collected by NMFS and WDFW were combined into a single data set to facilitate comparisons among individual studies (Appendix A).
In 1994, the U.S. Fish and Wildlife Service (USFWS) and Idaho Department of Fish and Game (IDFG) collected samples of steelhead from a number of natural populations in the Clearwater and Salmon River Basins in an attempt to determine whether releases of hatchery fish had affected the genetic structure of natural populations. Genetic analysis of these samples performed by NMFS (Waples 1995) indicated that none of the populations in the lower Clearwater River show evidence of substantial genetic introgression by steelhead from Dworshak NFH, in spite of widespread outplanting in the area. The samples from the Salmon River provided little clear insight into population structure. Two factors may have contributed to this latter result: 1) Some of the Salmon River samples were small (about 25 fish), thus limiting power to detect population structure, and 2) the populations sampled were among those believed most likely to have shown the effects of artificial propagation, so they may not be a good indication of native population structure.
In order to obtain a more complete picture of genetic structure of steelhead in California, NMFS worked with the California Department of Fish and Game to identify natural populations of O. mykiss that could be sampled without placing local populations at undue risk. Through these efforts, 10 samples (generally of 40-60 juvenile fish per sample) were collected and analyzed by NMFS for allozyme variation. Tissue samples from these collections are also being provided to J. Nielsen for use in her continuing studies of DNA variation in California O. mykiss. In addition, we analyzed four samples of steelhead from southern Oregon and northern California that were collected (but not analyzed) in 1992 as part of the status review for Illinois River winter steelhead. Important results from these new allozyme analyses can be summarized as follows.
1) The California samples show levels of population differentiation that are unprecedented for the species. At one locus (FBALD-3*; Fig. 2), a sample from the Klamath River was fixed for one allele and a sample from Gaviota Creek (near Santa Barbara) was fixed for another allele. A fixed allelic difference between populations is a rare occurrence for any Pacific salmon species, being generally encountered, if at all, only between populations at extreme ends of a geographic range. Previously, Busby et al. (1994) found a sharp transition in allelic frequencies at FBALD-3* in populations south of the Klamath River drainage, but their study did not include samples from south of the Eel River.
2) More detail about population structure of California steelhead can be obtained by examining Figures 3 and 4, which are different ways of summarizing patterns of genetic relationships based on Nei's (1978) unbiased genetic distance values between each pair of populations. Figure 3 is a dendrogram constructed using the unweighted pair-group method analysis (UPGMA) with arithmetic averaging, and Figure 4 is a different representation of the same data using multidimensional scaling (MDS). Multidimensional scaling plots allow one to view in two or three dimensions the pattern of relationships among populations; in contrast, a dendrogram is essentially a one-dimensional representation of the data. In general, two-dimensional MDS plots result in less distortion of the relationships among populations than do dendrograms, and three-dimensional plots have less distortion than two-dimensional plots. However, complex three-dimensional analyses are often difficult to represent in two-dimensional figures, so two-dimensional MDS plots are sometimes preferred for data sets that involve a large number of samples.
The new samples from the Chetco River and the Trinity River Hatchery and Cole Rivers Hatchery (Rogue River) cluster with samples previously analyzed from the Klamath Mountains Province, but the sample from Iron Gate Hatchery (Klamath River) is somewhat of an outlier. A new sample from the Middle Fork of the Eel River showed an affinity to the Mad River/Eel River/Redwood Creek group identified by Busby et al. (1994). The other new California samples were all quite different from any samples of coastal or inland steelhead previously examined. The sample from Coleman NFH and those from Mill and Deer Creeks (the two natural populations in the Sacramento River Basin believed to contain the most likely remnants of native steelhead) form a small, coherent group that is quite distinct from all other California steelhead. The remaining California samples (from Ten Mile River in Mendocino County to Gaviota Creek and Arroyo Hondo in Santa Barbara County) formed a cluster that diverged from the other samples at a genetic distance (Nei's D ÷ 0.03) higher than that previously found between coastal and inland races of steelhead (e.g., Busby et al. 1993).
In spite of the large interpopulational genetic differences, the pattern of population structure south of the Eel River is not entirely clear from the allozyme data. The MDS plot (Fig. 4) shows the magnitude of the diversity but also illustrates the difficulty in drawing inferences about geographic population structure. For example, the two samples from Santa Barbara County are genetically quite divergent but are not similar to each other. In the allozyme analysis, we found only modest genetic differences between the samples from Ten Mile River and Lagunitas Creek (a tributary of Tomales Bay), both of which are north of San Francisco, but we also found that these samples were more similar to the sample from Whale Rock Hatchery (near San Luis Obispo) than they were to the geographically closer samples from the San Francisco-Monterey area (Scott Creek, Carmel River, and San Lorenzo River).
3) Our allozyme analysis is consistent with Nielsen's (1994) mtDNA study in finding a high degree of interpopulational differentiation within California. The allozyme data also support Nielsen's finding of large genetic differences among samples from southern California. One notable difference in the two analyses is that, whereas Nielsen found substantial differences in the frequencies of some mtDNA alleles between samples from Mendocino and Marin Counties, we did not (as evidenced by the relative similarity between the Ten Mile River and Lagunitas Creek allozyme samples).
4) To examine possible explanations for the distinctive genetic characteristics of the samples from the Sacramento River, we performed another analysis that included data for five rainbow trout samples provided by WDFW, including the four that Phelps et al. (1994) used in their comparisons with Washington state steelhead. This new analysis (Figs. 5 and 6) showed that steelhead from the Sacramento River are genetically more similar to these rainbow trout populations than they are to coastal steelhead populations from California. This could be the result of integration of rainbow trout into the steelhead broodstock at Coleman NFH and subsequent effects of Coleman NFH strays or outplants on natural populations in the Sacramento River; the difficulty of distinguishing between nonanadromous and anadromous O. mykiss during broodstock collection in the upper Sacramento River has been described by several authors (e.g., Hallock et al. 1961, Behnke 1992, Cramer et al. 1995). On the other hand, this genetic similarity could simply reflect ancestral relationships, since the origins of most of the present-day rainbow trout stocks can be traced to collections of anadromous and nonanadromous O. mykiss from the McCloud River in the Sacramento River Basin made early in the century (Behnke 1992).
To facilitate direct comparisons among recent collections by WDFW (whose samples are all from within Washington state) and NMFS (whose new data are primarily from the Snake River, Oregon, and California), the two agencies collaborated to integrate their genetic data for steelhead into a single data set. Lead scientists for this collaboration were Stevan Phelps from WDFW and Paul Aebersold from NMFS. Extensive inter-laboratory communication that included exchange of recipes and procedures and detailed review of photographic records indicated a high degree of consistency in data for the two agencies.
To reduce the number of samples in the analysis, collections from different years within locations were combined to form single pooled samples for each location. This combined data set includes information for samples from 100 natural populations of steelhead and rainbow trout and 3 steelhead hatcheries. To account for some differences between laboratories in the suite of gene loci examined, the number of polymorphic gene loci used in this combined data set (42) was slightly less than the full complement used by either NMFS or WDFW in their individual analyses. Therefore, those individual studies provide more detailed information about genetic relationships within particular geographic areas. Nevertheless, the combined data set provides the broadest geographic coverage to date for a data set that takes advantage of significant advances in recent years in the number of genetic markers for O. mykiss. Results of analyses of this data set are shown in Figures 7 and 8.
These figures show clearly the major difference between coastal and inland forms. The position of one of the Klickitat River samples, #94 (Bowman Creek), which is not clearly aligned with either the inland or coastal groups in Figure 8, may reflect the genetic influence of hatchery coastal rainbow trout overlaid upon a native inland steelhead population (Phelps) . Within the inland group, samples from the middle Columbia River form a separate subgroup from those in the Snake River, with one Yakima River Basin sample (Toppenish Creek) being an outlier. No recent genetic data are available for populations in Columbia River tributaries upstream from the Yakima River. Dworshak NFH was included in these figures to provide an indication of the relative distinctiveness of this population.
Within the coastal group, the geographic differences among Washington populations detected by Leider et al. (1995) are modest in comparison to the overall pattern of diversity from Puget Sound to southern California. However, there are a number of coastal steelhead samples from Washington that do not show a strong genetic affinity to any other populations. The most notable outliers in the combined analysis were samples from the Upper Chehalis, Washougal, Nooksack, and Tahuya Rivers. With the benefit of this combined data set, it is apparent that populations south of Cape Blanco are genetically distinct from all northern populations. Previous analyses that considered steelhead from the Klamath Mountains Province (Busby et al. 1993, 1994) did not include samples from Washington, so this was the first time a direct comparison of the two groups has been possible.
The relative magnitude of genetic diversity among steelhead populations from California is readily apparent from these figures. In the dendrogram, California samples from south of the Eel River form a genetically diverse cluster that joins the other west coast steelhead populations external to the inland-coastal break in Washington and Oregon. Although California steelhead, including those in the Sacramento River, are more similar genetically to other coastal steelhead populations than they are to inland steelhead from the Columbia River Basin (see Fig. 8), the genetic diversity within the coastal steelhead lineage is considerable.
The old and new genetic data from allozyme, DNA, and chromosomal studies can be synthesized and summarized as follows: 1) All studies that have addressed the question have found large genetic differences between coastal and inland forms (both anadromous and nonanadromous) of O. mykiss. In the Columbia River Basin, the boundary between the two forms occurs at approximately the Cascade Crest. However, available data are not sufficient to determine with certainty whether there is a transitional (perhaps intergrade) zone between the forms or whether they remain discrete. If the two forms are discrete even in this area of closest contact, the exact boundaries are unclear.
2) Genetic data do not support the hypothesis that winter- and summer-run steelhead are separate monophyletic units, as appears to be the case with inland and coastal forms. Rather, steelhead with different run timing in the same geographic area may be genetically more similar than either is to fish from another area with a similar run timing. This result, however, does not mean that there cannot be genetic differences between summer and winter steelhead in any given drainage.
3) The inland steelhead lineage is represented only by populations in the Columbia and Fraser River Basins. Within the inland group, consistent differences are found between populations from the Snake and Columbia Rivers, and there is also evidence for a modest level of population differentiation between major drainages within each of these two rivers. Steelhead from Dworshak NFH are genetically the most distinctive population within the inland lineage.
4) Within the geographic area covered by this status review, coastal steelhead occur in a diverse array of populations. A large group showing consistent geographic structure but relatively modest genetic differences between populations includes most samples from Puget Sound, coastal Washington, and the lower Columbia River. The few recent samples from coastal Oregon north of Cape Blanco show some differences from this larger group, and populations from the Klamath Mountains Province are genetically different from those to the north or south. South of the Klamath River, large differences are found between coastal populations, both for allozymes and DNA markers, but the geographic structure to the variation is not fully resolved. Samples from the Sacramento River are quite distinct from all other coastal and inland populations that have been sampled but show some affinity to hatchery rainbow trout.
Any of several factors might explain the much higher diversity among coastal steelhead populations in California than in Washington. First, it is possible that diversity among populations in Washington was greater historically but has been eroded by human influence. There is a long history of widespread releases of a few hatchery stocks in Washington (see Reisenbichler and Phelps 1989). Phelps et al. (1994) found evidence for substantial effects of hatchery stocks on several winter steelhead populations. The occurrence of a number of quite distinct natural populations of coastal steelhead (e.g., upper Chehalis, Washougal, Wind, Nooksack, and Tahuya Rivers) might be explained if they represent remnants of more complex population structure that occurred historically.
Two factors, either individually or in combination, likely contributed to the high level of genetic variation in California steelhead. First, environmental conditions in most streams in central and southern California are extreme for anadromous salmonids, with low flows and summer water temperatures that may reach or exceed typical thermal limits for O. mykiss. These environmental conditions may promote strong local adaptations and inhibit gene flow among populations. In addition, sand berms at the mouth of streams may completely block migration of juveniles or adults in some years, and it is likely that this phenomenon fosters flexible life history patterns, a greater importance of resident fish, and increased opportunities for isolation. Second, it is possible that these factors, which historically would be likely to promote isolation and differentiation, have been intensified by human-mediated events of the past century. Major water projects in California have reduced stream flows and increased the frequency and duration of stream blockages. South of San Francisco Bay, a high proportion of native steelhead populations are declining or already extinct (Titus et al. in press), indicating reduced opportunities for genetic contact between populations. In addition, declines in abundance in the populations that do remain facilitate more rapid differentiation among populations due to genetic drift.
Based on a review of the biology and ecology of west coast steelhead, the Biological Review Team (BRT) identified 15 ESUs, 12 of which include coastal forms and 3 of which include inland forms (Fig. 9, Appendix B). Genetic data (from protein electrophoresis, DNA markers, and chromosomal analysis) were the primary evidence considered for the reproductive isolation criterion, supplemented by inferences about barriers to migration created by natural geographic features. A number of factors were considered to be important in evaluations of ecological/genetic diversity, with data for migration and spawn timing, life history, ichthyogeography, hydrology, and other environmental features of the habitat being particularly informative. In the following summaries, we describe only those factors that were valuable in making individual ESU determinations.
Each of the ESUs includes multiple spawning populations of O. mykiss, and most ESUs also extend over a considerable geographic area. This result is consistent with NMFS' species definition paper, which states that, in general, "ESUs should correspond to more comprehensive units unless there is clear evidence that evolutionarily important differences exist between smaller population segments" (Waples 1991b, p. 20). However, considerable diversity in genetic or life history traits or habitat features may exist within a single complex ESU, and the descriptions below briefly summarize some of the notable types of diversity within each ESU. This diversity is considered in the next section in evaluating risk to the ESU as a whole.
Evolutionary Significance of Phylogenetic and Life History Forms
In defining ESUs for west coast steelhead, it is necessary to consider the significance of the various phylogenetic and life history forms that have been described (see page 7). Of these, only the coastal or inland groups are considered monophyletic, the others apparently being adaptive expressions of the life history plasticity that is characteristic of O. mykiss. Genetic analyses show large and consistent differentiation between coastal and inland steelhead. Also, these groups have not been found to co-occur in nature, although some degree of overlap in the Columbia River Basin in the vicinity of the Cascade Crest cannot be ruled out.
Unlike the coastal/inland groups, summer and winter steelhead co-occur in several river basins, primarily within the range of the coastal steelhead group. The few genetic analyses that have considered this issue indicate that summer and winter steelhead from the same river basin are more genetically similar to each other than to the same run type in another river basin. This indicates that all summer steelhead, for example, are not descended and distributed from one ancestral source and, therefore, are not a monophyletic unit.
Genetic assessments within and between river basins have not specifically been conducted on A- and B-run steelhead. One reason for this is that most genetic analyses in the Snake River Basin have been conducted on juvenile steelhead, while the characteristics of A- and B-run are manifested in the adults. We do know that there is geographic structure to the genetic data that does not appear to strictly follow the distribution of the A- and B-runs within the Columbia and Snake River Basins.
Half-pounders are only reported in the literature from a small geographic region in southern Oregon and northern California. However, genetic data do not show a particularly strong affinity among rivers having half-pounders; rather, the affinities are geographic, including streams both with and without half-pounders. Additionally, winter steelhead broodstock for Cole Rivers Hatchery on the Rogue River in Oregon were initially selected for fish without evidence of the half-pounder life history, yet there is evidence that among winter steelhead subsequently returning to the hatchery, approximately 30% underwent a half-pounder migration (Evenson) , suggesting that this is not strictly a genetic trait.
Resident fish--Few detailed studies have been done on the relationship between resident and anadromous O. mykiss in the same location. Genetic studies generally show that the two forms from the same area are more similar to each other than either is to the same form from a different geographic area. Thus, rainbow trout and steelhead from the same area may share a common gene pool, at least over evolutionary time periods. It is also generally believed (although definitive information on this topic is scarce) that progeny of nonanadromous O. mykiss can be anadromous, and that anadromous O. mykiss can produce nonanadromous progeny. It was the consensus of fishery biologists we consulted throughout the region that resident fish should generally be considered part of the steelhead ESUs. On the other hand, there is also evidence for substantial genetic divergence between resident and anadromous fish in areas where resident populations have been isolated by long-standing natural barriers. In addition, hatchery rainbow trout derived from a few mixed strains have been widely planted throughout the range of west coast steelhead, and resident populations established as a result of such transplants would not be native to the area.
Based on these considerations, the BRT concluded that, in general, the ESUs described below include resident O. mykiss in cases where they have the opportunity to interbreed with anadromous fish. Geographic areas in which the role of resident fish may be particularly important include southern California and the upper Columbia River; in both of these areas, extreme environmental conditions may promote increased flexibility in life history strategies for native populations of O. mykiss. Resident populations above long-standing natural barriers, and those that have resulted from the introduction of non-native rainbow trout, would not be considered part of the ESUs. Resident populations that inhabit areas upstream from human-caused migration barriers (e.g., Grand Coulee Dam, the Hells Canyon Dam complex, and numerous smaller barriers in California) may contain genetic resources similar to those of anadromous fish in the ESU, but little information is available on these fish or the role they might play in conserving natural populations of steelhead. The status with respect to steelhead ESUs of resident fish upstream from human-caused migration barriers must be evaluated on a case-by-case basis as more information becomes available.