The post-Wisconsinan glacial biogeography of bull trout ... › rm › boise › publications...rately on the truss and meristic characters (e.g., Thorpe 1976) sug-gest that there

The post-Wisconsinan glacial biogeography of bulltrout (Salvelinus confluentus): a multivariatemorphometric approach for conservation biologyand management

Gordon R. Haas and J.D. McPhail

Abstract: Canonical correlation analysis (CCA) can quantitatively partition historical and ecological information frommorphometric data where these features are otherwise confounded. CCA is applied to sample site localitymorphometric data and corresponding sample site locality coordinate data for bull trout. Two vectors result. The firstaccounts for the maximum morphometric variation correlated to geographic information specified by the localitycoordinates. The second represents the remaining less correlated variation. For biogeography, the first vector generateshistorical hypotheses for Pleistocene glacial refugia and for post-Wisconsinan glacial recolonization patterns andphylogeographic relationships. The second vector infers hypotheses for broad ecological patterns. The historicalbiogeographic patterns for bull trout suggest recolonization from either two or three glacial refugia and emphasizewithin-species biodiversity in western North America. These patterns from the Chehalis and Columbia refugia arelargely concordant with other analyses based on molecular genetics. The morphometric analysis also suggests theadditional possibility of a Nahanni and (or) Bering refugium. The ecological patterns suggest the importance and extentof anadromy and migration within these historical groups and how this may have affected postglacial recolonization,present distributions, and life histories.

Résumé : L’analyse des corrélations canoniques (CCA) peut séparer de façon quantitative les informations historiques etécologiques des données morphométriques, lorsque celles-ci sont entremêlées. Une analyse CCA menée sur des donnéesmorphométriques de l’Omble à tête plate reliées à des sites d’échantillonnage ainsi que sur les données du milieucorrespondant aux mêmes sites a généré deux vecteurs. Le premier vecteur représente le maximum de variation morpho-métrique en corrélation avec les données géographiques fournies par les coordonnées du milieu. La second vecteur illustrele reste de la variation moins corrélée. En ce qui a trait à la biogéographie, le premier vecteur génère des hypothèseshistoriques sur les refuges glaciaires pendant le Pléistocène, sur les voies de recolonisation après les glaciations duWisconsinien et sur les relations phylogéographiques. Le second vecteur génère des hypothèses de portée écologique gé-nérale. La structure biogéographique historique laisse croire à l’existence d’une recolonisation à partir de deux ou trois re-fuges glaciaires pour l’Omble à tête plate et met en évidence la variation de la biodiversité intraspécifique dans l’ouest del’Amérique du Nord. Ces structures associées aux refuges de Chehalis et de Columbia s’accordent en grande partie avecles résultats d’analyses basées sur la génétique moléculaire. L’analyse morphométrique laisse croire à la possibilité d’unrefuge additionnel dans le Nahanni et (ou) la Béringie. Les structures écologiques démontrent l’importance et l’étendue dela migration et des déplacements anadromes chez ces groupes historiques et illustrent comment ils ont pu affecter la reco-lonisation post-glaciaire, les répartitions géographiques actuelles et les cycles biologiques.

[Traduit par la Rédaction] Haas and McPhail 2203

Introduction

Biogeography is the study of the distribution of organismsand their variability in space and time. This discipline isoften used to discern groupings of animals for conservationand management purposes such as within the Endangered

Species Act in the United States (U.S.) (e.g., fish, seeWaples 1995). Biogeography is seemingly split into twoapproaches with different aims and time scales (e.g., Ball1975; Endler 1982a; Birks 1987). The first, the ecologicalschool, studies the dispersion of organisms and the mecha-nisms and environmental interactions that maintain or

Can. J. Fish. Aquat. Sci. 58: 2189–2203 (2001) © 2001 NRC Canada

2189

DOI: 10.1139/cjfas-58-11-2189

Received June 7, 2000. Accepted August 27, 2001. Published on the NRC Research Press Web site at http://cjfas.nrc.ca onNovember 5, 2001.J15802

G.R. Haas1,2 and J.D. McPhail. Centre for Biodiversity Research, Department of Zoology, and Native Fish Research Group,University of British Columbia, 6270 University Boulevard, Vancouver, BC V6T 1Z4, Canada.

1Corresponding author (e-mail: [email protected]).2Present address: School of Fisheries and Ocean Sciences and University of Alaska Museum, University of Alaska Fairbanks,245 O’Neill Building, Fairbanks, AK 99775-7220, U.S.A.

change this dispersion (e.g., MacArthur and Wilson 1967).This research is usually done at the population or commu-nity level and sometimes involves direct experimentation.The second is the historical school that studies spatial andtemporal distribution patterns (e.g., Brooks 1985; Wiley1988). This work is conducted at the taxonomic level and at-tempts to explain distributions and interactions using pastevents. Direct experimentation is thus not generally possibleand explanations are usually inferential (Gould and Wood-ruff 1990).

The ecological school focuses on local populations inwhich diversity often has developed within a short evolution-ary time scale. These researchers are usually not interestedin broad biogeographic or species-specific patterns and con-sequently often ignore the potential influence of historicalevents and phylogeny. The historical school finds small-scale ecological differentiation problematic. Their tendencyis to work at a higher taxonomic level and look for generalbiogeographic patterns. Historical analyses usually do notdeal with subspecific or localized variation or with bio-

geographic patterns of single species (Brooks 1985; Cracraft1988; Gorman 1992).

The approaches of the ecological and historical schools tobiogeography are thus generally separate and confounded. Acomplete biogeographic analysis for conservation or man-agement purposes really should contend with the integrationof both ecological and historical information (e.g., Endler1982a, 1982b; Mayden 1992). A combined procedure shoulddiscern which data aspects are related to these categories andwhat it says about them singly and in relation to each other(Cracraft 1988). This lack of congruence can result in diffi-culty or bias in the rigorous analyses of unconfounded com-plete biogeographic patterns, especially for single species.

Canonical correlation analysis (CCA) could be used in acombined ecological and historical approach within bio-geography. CCA can quantitatively separate confounding in-formation, uses multivariate data sets, and operates at alllevels. For our biogeographic analyses, CCA essentiallyquantifies the multivariate relationship between a set ofmorphometric and corresponding locality coordinate matri-

© 2001 NRC Canada

2190 Can. J. Fish. Aquat. Sci. Vol. 58, 2001

Fig. 1. The distribution of bull trout (open circles) in the conterminous United States for this study as sampled in Haas and McPhail(1991). The triangles represent bull trout site localites of sympatry with Dolly Varden that are now known but for which collectionswere not permitted and samples were unavailable for our original research, and thus were also not analyzed for this study. For samplesite locations, see corresponding numbers in the Appendix.

ces. The maximum morphometric variation correlated to thegeographic information specified by the locality coordinatesgenerates hypotheses for historical biogeographic patterns.The remaining variation not as strongly correlated or evenuncorrelated to geography is used to infer hypotheses ofbroad ecological biogeographic patterns. These historicaland ecological features could then be integrated and exam-ined in relation to each other for a complete and detailedbiogeographic analysis.

To demonstrate, we use CCA to analyze the biogeographyof bull trout (Salvelinus confluentus) throughout its broadrange (Haas and McPhail 1991). Fish are ideal for bio-geographic study because they are restricted to living in wa-ter and their distribution patterns must thus reflect specificdiscernible aquatic connections. Bull trout are also longrenowned for their plasticity, which has previously con-founded attempts to isolate historical or ecological patternsin their morphology (e.g., McPhail 1961; Savvaitova 1980).Bull trout are endemic to and distributed throughout much ofwestern North America, particularly in its interior regions(Figs. 1 and 2). This area was strongly and differentiallyaffected by Pleistocene glaciation (Lindsey and McPhail

1986), with char (Salvelinus spp.) being well recognized fortheir rapid postglacial recolonization (Balon 1984; Milner etal. 2000; Oswood et al. 2000). Bull trout often exhibit strongmigrations and their life history is usually summarized intofour migratory types: resident, fluvial, adfluvial, and anadro-mous or at least sea-run. The genetic variability of bull troutis usually low within populations and often marked betweenthem (e.g., Leary et al. 1993; Taylor et al. 1999), and ametapopulation model has consistently been invoked toaccount for this, at least in watersheds on a locally intercon-nected scale (e.g., Dunham and Rieman 1999; Spruell et al.1999). These morphological and genetic patterns of variabil-ity for bull trout, combined with the biogeographic complex-ity of western North America, should appropriately test theability of CCA to partition historical and ecological variation.Finally, other biogeographic analyses of bull trout usingmolecular genetic data are available (Bellerud et al. 1997;Spruell and Allendorf 1997; Taylor et al. 1999) to provide acontext for and test of this morphometric procedure.

Furthermore, such biogeographic information is vital be-cause bull trout are listed as a threatened species throughoutthe conterminous U.S. under their Endangered Species Act

© 2001 NRC Canada

Haas and McPhail 2191

Fig. 2. The distribution of bull trout (open circles and triangles) in Canada for this study as sampled in Haas and McPhail (1991). Forsample site locations, see corresponding numbers in the Appendix.

and as a species of special concern in most of their Canadianrange. Conservation and management programs attemptingto deal with their variability must recognize the interplay ofhistorical and ecological information. The variabilitiesevolved through long-term historical or short-term ecologi-cal events can be very different and are both worthy of pres-ervation. These biogeographic events could also furthersuggest different susceptibilities to perturbation or extinctionand may help prioritize conservation efforts if that becamenecessary.

Materials and methods

Data sets

Morphometrics (morphology and meristics)Twenty-three truss measurements were computed from 11 land-

mark points (e.g., Bookstein et al. 1985). Two meristic variables,total anal fin and branchiostegal ray numbers, were scored. Alltruss and other data collection was consistently done by G.R. Haasand is described in full detail in Haas and McPhail (1991).

Measurement error was assessed by remeasuring five fish. Themean amount of measurement error is negligible. A one-way anal-ysis of variance (ANOVA) reveals that the ratio of individual to

total (among plus within) variation is large (close to one), indicat-ing that measurement repeatability is high and measurement erroris insignificant (Falconer 1981).

Sexual dimorphism was tested using 30 fish. Two-way ANOVAand two-way multiple analysis of variance (MANOVA) done sepa-rately on the truss and meristic characters (e.g., Thorpe 1976) sug-gest that there is no significant univariate or multivariate sexualdimorphism (p ≥0.1). Furthermore, principal component analysesdone separately on the truss and meristic variables reveal no differ-ences between males and females (Haas 1988). Homoscedasticitywas conservatively tested using Box’s modification of Bartlett’stest (e.g., Pimentel 1979). Our log10-transformed data are homo-scedastic for both sexes. The data could therefore be analyzed intotal.

Localities and their coordinatesThe 693 bull trout examined are from 171 sample site localites

throughout their North American range (Figs. 1 and 2). The spe-cific locality names are given in the Appendix. The locality coordi-nates used are the latitudes and longitudes of the sample sites.Localities were estimated when it was not possible to get an accu-rate site locality specification because of insufficient museum cata-logue information. These estimates are based on most probableaccess to a site locality or were put at its centre. All locality coor-dinates were taken to minutes, and these minutes were converted to

© 2001 NRC Canada


Fig. 3. Dendrogram showing the hypothesized post-Wisconsinan historical biogeographic relationships of bull trout. The site locality num-bers given are explained in the Appendix. The watersheds and theorized glacial refugia summarized by each node are listed in turn belowthem. If the summarized region is spread out over two nodes, so is its descriptive name, which is also in square brackets. These hypothe-sized historical biogeographic relationships of bull trout match those for molecular genetic data except in the most northwestern portion oftheir range for which the genetic interpretation is presented in braces. Explained errors are site-locality misrepresentations (see Resultssection) that nonetheless do not directly fit these general distributions and are marked with a open triangle. Unexplained errors aremistakes in site localities that cannot be readily accounted for and are designated with a solid triangle. Their error rates given are thenumber of respective triangles divided by the number of localities, and the total overall error rate is the same but for the total number ofopen plus solid triangles.

decimals for the statistical analyses. Longitudes are not equidistantthroughout the sampled ranges, but our mathematical adjustmentfor this (Snyder 1983) did not alter results.

Statistical analyses

Locality means—biogeographic summarizationThe original morphometric and locality data sets were independ-

ently reduced to locality means prior to all statistical analyses. Thiswas done because it is localities, not individuals, that are beingbiogeographically examined here and to necessarily compensatefor discrepancies in site locality sample sizes. This data reductionto locality means could have reduced or increased the amount ofrelated variability at each locale. However, this approach was nec-essary and simple, and an average was the most suitable measureof central tendency to use given the distribution of this data. Therewas no evidence that such population variability problems existedbecause other measures of central tendency and subsequent jack-knife trials (more fully described later in Results) provided thesame overall results.

Principal component analysis (PCA)—allometric adjustmentPCA was performed on the total covariance matrix of log10-

transformed locality means of the 23 truss measurements. Log-arithmic transformation of morphological data is usually recom-mended (Bookstein et al. 1985; Shea 1985), and here it improved

univariate and multivariate normality and helped standardize char-acter variances and linearity (Haas and McPhail 1991). Even still,a PCA performed on raw data gave similar results. Our univariatenormality test was the probability (quantile) plot correlation coeffi-cient procedure (e.g., Filliben 1975). Our multivariate normalitytest was also based on probability plots and allowed for a visualassessment of whether the probability plot approximated a straightline that would represent multivariate normality (e.g., Andrews etal. 1973). As well, our PCA scatter plot has an ellipsoidal distribu-tion (Haas and McPhail 1991) characteristic of multivariately nor-mally distributed data (Thorpe 1976; Reyment et al. 1984).

Only our first two eigenvectors and principal components (PCs)are significant and together account for about 98% of the variance.The first PC represented about 97% and the second PC representedabout 1% of this variance. Significance was tested using Bartlett’sχ2 test of sphericity (Pimentel 1979) and was further corroboratedby the Scree test. We used only the second PC in the canonical cor-relation analysis because it contained and summarized all of thestatistically significant shape information statistically independentof (orthogonal to) size variation (also see Haas and McPhail 1991).

The first PC was interpreted here as, and is traditionally, a “sizeinformation” component, and the second PC was interpreted as a“shape information” component. The classic morphometric PCAfeatures for designating this size and shape are present (Haas 1988;Haas and McPhail 1991) and suggest that allometric size waseffectively removed in the first PC (e.g., Reyment et al. 1984). The

© 2001 NRC Canada


Fig. 4. Dendrogram showing the hypothesized ecological biogeographic relationships of bull trout. The site locality numbers given areexplained in the Appendix. The watersheds and theorized ecological patterns of anadromy and migration summarized by each node arelisted below them. If the summarized region is spread out over two nodes, so is its descriptive name, which is also in square brackets.The watershed names in braces are those that do not as closely or completely fit the overall ecological representation. Explained errorsare site-locality misrepresentations (see Results section) that nonetheless do not directly fit these general distributions and are markedwith a open triangle. Unexplained errors are mistakes in site localities that cannot be accounted for and are designated with a solid tri-angle. Their error rates given are the number of respective triangles divided by the number of localities, and the total overall error rateis the same but for the total number of open plus solid triangles. The circled node represents a division not easily interpreted here.

first eigenvector used to calculate our first PC is general (all load-ings have the same sign) and accounts for the vast majority of thevariability (large eigenvalue). Its general sign is indicative of a sin-gle type of variation (Pimentel 1979) and its large eigenvaluesmake intuitive sense because size-related variation predominates.Our second eigenvector is bipolar (loadings have mixed signs),which is typical of a shape component. Further support for our sizeand shape interpretation is that bull trout length and weight arestrongly correlated with PC1 and uncorrelated with PC2.

This PCA allometric adjustment was corroborated using a regres-sion technique for calculating a mean shape individual for each local-ity (e.g., Shea 1985). This regression is based on log10-transformeddata and uses the slopes derived from the regression to adjust a set ofvariables for each locality to the overall grand mean body size for abull trout sample. This regression shape data produced comparable re-sults to the PCA (Haas 1988) and, when used in the canonical corre-lation analysis, performed similarly to the PCA shape data.

The two meristic variables used are not size dependent (Haasand McPhail 1991) and were analyzed as locality means of theirlog10-transformed data. This was done in conjunction with thesecond PC morphological shape vector to create a final size-independent morphometric data matrix. These meristic variableswere log10 transformed as it made their variances and distributionsmore homoscedastic and univariately–multivariately normal andhelped standardize their linearity, variability, and scale. Their log10transformation did not substantially alter the results but did helpthe morphometric matrices conform more clearly to multivariatestatistical assumptions (Gittins 1979).

Canonical correlation analysis (CCA)—historical–ecological biogeographic analyses

CCA was applied to a matrix consisting of the morphologicalshape PC2 scores and the two meristic variables and to a matrix of

© 2001 NRC Canada


Fig. 5. (a) Map of western North America detailing the present jurisdictions and main watersheds–drainages in the entire known rangeof bull trout. (b) Map of western North America showing our hypothesized post-Wisconsinan glaciation recolonization routes for bulltrout. The solid arrows indicate their most hypothesized routes and these match those derived from molecular genetic data (Taylor etal. 1999). Different-shaded solid arrows differentiate the Chehalis and Columbia refugia. The striped arrows represent hypothesizedrecolonization from the Bering or Nahanni refugia, but these morphometric patterns are not supported by the same molecular geneticdata set as depicted here by the dotted arrow (see Discussion section). The heavy internally hatched lines represent the maximum ex-tent of Wisconsinan glaciation, and the dotted lighter grey arrows show the relatedness and possible connections of the drainages lessor unaffected by it.

© 2001 NRC Canada


the corresponding locality coordinates. The significance of the twovectors resulting from the CCA was tested using a χ2 test of sphe-ricity (e.g., Tabachnik and Fidell 1989), which tests whether eachvector is statistically significant by examining if there is significantoverlap in the variability between the two matrices entered into theCCA.

CCA correlation coefficients and redundancy coefficients(Green 1978; Pimentel 1979; Wartenberg 1985a) were used to as-sess how much of the variation is accounted for. The CCA correla-tion coefficients are measures of the correlation between the sets ofcanonical vectors corresponding to the two original data matrices.Redundancy coefficients quantify the amount of variance of onedata matrix explained by the new CCA vector derived for the otherdata matrix.

A jackknife technique (N = 100) was applied to the CCA andsubsequent cluster analyses to assess robustness and to check forspurious correlations and relationships (Gittins 1979; Chernoff1982). Localities and their corresponding morphometric data weresequentially and randomly removed from the data matrices beforethe CCA. The number of localities jackknifed varied, except thatno complete drainages as set out in the Appendix were removed.The effect of such a complete drainage basin removal would be in-valid because it could obscure biogeographic relationships basedon previous watershed associations involving that watershed.

Canonical trend surface analysis (Wartenburg 1985a) is an ana-lytical technique similar to CCA (e.g., Gittins 1979) and was usedas a final test. It produced the same results. CCA was then used be-cause it is more direct and is readily available in most statisticalsoftware for potential future use of such methodology.

Cluster analyses—representation of canonical correlationanalysis (CCA) results

The two resulting vectors from the CCA were independently en-tered into a cluster analysis. Standard unweighted average linkagecluster analysis based on Euclidean distances was used. This is of-ten recommended as being a more natural representation (Pimentel1979; Pielou 1984), and other coefficients did not overly affect ourresults (Birks 1987; Hughes et al. 1987). The dendrograms aremainly used for more informative and intuitive graphic representa-tion, and plots of the CCA values reveal clusters that match thedendrogram branchings.

The dendrograms are given three rates of percentage error: ex-plained error, unexplained error, and total overall error. Explainederrors are not true errors and represent explicable locality misre-presentations usually involving isolated sample site localitiesdistant from their true watershed grouping and close to a set ofsample site localites in the neighbouring watershed grouping. Un-explained errors are defined as clustering mistakes that cannot bereadily explained and basically are true errors. The unexplainederror rate is the number of populations with unexplained errorsdivided by the number of localities. The total overall error rate isthe number of explained plus unexplained errors divided by thenumber of localities.

The dendrogram patterns were tested to determine if the CCAwas simply creating patterns based on geographic proximity ornearest neighbours. A minimal spanning tree (MST) analysis (e.g.,Gower and Ross 1969) was used to statistically establish the short-est possible connections between sample site localities and wascompared to the CCA dendrograms. An MST statistically connectsall geometric vertices of the locality data to keep the sum of alltheir edge lengths as small as possible, whereas CCA and our clus-ter analyses make no such a priori calculations.

Results

Historical biogeographyThe first CCA vector portrayed in the cluster analysis cor-

responds closely to possible hypothesized historical bio-geographic patterns for bull trout (Fig. 3). Furthermore,these CCA patterns do not fit the shortest possible connec-tions between, or the resultant group patterns for, the samplesite localities as calculated by MST analyses. Localities neareach other are not necessarily closest in affiliation so thisvector is not solely based on geographic proximity.

The higher canonical correlation and redundancy coeffi-cients for these first vectors (Fig. 3) compared with the sec-ond vectors (Fig. 4) indicate that more morphometric dataare correlated with the locality information. However, this isat least partly the natural result of the CCA statistically ac-counting for more variability in the first vector. These ca-nonical correlation and redundancy coefficients are not verylarge, but there is no general rule for establishing lower lim-its (Pimentel 1979). These low coefficients may be the resultof correlating such disparate features as morphometrics andlocality coordinates over a large geographic area. One mightnot expect much statistical association in terms of redun-dancy between these two features, but the resultingmorphometric patterns correlated to this geography can stillmake sense. The first vector is statistically significant atp < 0.001, and the relationships were very robust in jack-knife trials. An interpretation of these first vectors shouldthus be statistically acceptable (Green 1978; Wartenberg1985a), and the historical biogeographic hypotheses basedon them should be reliable (Tabachnik and Fidell 1989).

There are low unexplained error and total overall errorrates for this dendrogram (Fig. 3). Seven sets of unexplainederrors and two sets of explicable error misrepresentations arefound. The two sets of misrepresentations are two samplesthat are located very close to the other drainages with whichthey were clustered. This could be because these samples areisolated by distance from other site localites in their drain-age groups, especially in the case of the relatively poorlysampled northern regions where one of these misrepresenta-tions occurred.

Ecological biogeographyThe second vector resulting from the CCA corresponds to

the broad ecological and biogeographic phenomena ofmigration and anadromy in bull trout (Fig. 4). Thebranchings and resultant group patterns are not solely basedon geographic proximity as can be seen on the dendrogramsand statistically through comparisons with the MST analysisresults. Migration is important because it likely had the larg-est effect on postglacial recolonizations and present distribu-tions. As well, migration and corresponding metapopulationshave been postulated as being very important in bull trout(e.g., Dunham and Rieman 1999; Spruell et al. 1999), andsuch migration patterns do help to explain some of theirunique morphometric and life-history situations.

The canonical correlation and redundancy coefficients arelower (Fig. 4) than for the first historical biogeography CCAvector (Fig. 3). Again, there is no general rule for establish-ing lower limits, and this second vector is statistically signif-icant (p < 0.015). This indicates that there should be areliable relationship between the morphometric and localitymatrices in the second canonical correlation for bull trout(Green 1978; Wartenberg 1985a; Tabachnik and Fidell1989). The relationships were also quite robust in jackknife

trials but not as strong as those for historical biogeographyin the first vector.

The low canonical correlation and redundancy coefficientsmay again be the result of correlating such disparate featuresas morphometrics and geography over a large area and of thefirst vector accounting for more variability as a naturalmultivariate statistical result. As well, these second vectorscontain any random variation and remaining historical infor-mation in the data sets leftover after the first canonical cor-relation. This could be contributing to any difficulties intheir interpretations.

Whatever historical data remain in these second ecologi-cal vectors are natural statistical consequences of the CCAbut are also likely contributing appropriate and useful infor-mation to ecology and migratory patterns. It would be mis-leading to argue that all historical information is noteffectively removed in the first vectors and that history andecology are thus still confounded. History and ecologyshould have complex interactions and not be totally andcompletely distinct statistically or in natural biological situa-tions (Endler 1982b; Cracraft 1988; Mayden 1992). A paral-lel and more common related example in multivariatestatistics is the use of PCA to partition size and shape intoits first two respective vectors (e.g., Haas 1988; Haas andMcPhail 1991). For similar statistical reasons as in the CCA,the second PCA shape vector still usually contains some sizeinformation, but this is generally viewed as contributing toshape and its relationships. The first PCA vector does effec-tively remove the necessary allometric size information.

There are two unexplained sets of errors and five explain-able sets of misrepresentations (Fig. 4). These misrepresen-tations may be quite minor because they are clustered on theother side of a minor division within an overall ecologicalbiogeographic group largely related to anadromy. The singleisolated “hanging” population (No. 308) is not easily ex-plained in terms of its unique clustering but it does stillgroup with the correct overall division and also is the mostnortherly and discretely isolated sample in this study. Thecircled node on the dendrogram branchings represents theonly division not readily explicable from an ecological view-point. None of the site localites here that had errors or mis-representations matchs those for historical biogeography.

Discussion

Historical biogeographyThe hypothesized historical biogeographic pattern for bull

trout could be interpreted as presently existing in or havingrecolonized Wisconsinan-glaciated North America from twoto three distinct refugia, with one of these further subdividedinto two distinct regions. The first split in the dendrogramlargely corresponds to the left branch, representing riverseast of the Continental Divide, and to glaciated or stronglyglacially affected portions of the Columbia River. The rightbranch largely represents rivers west of the Continental Di-vide and those south of the extent of Wisconsinan glaciation.The next level of splits in the dendrogram divides each ofthese two groups into two more biogeographic units.

The North and South Saskatchewan rivers and the extremeupper Columbia–Flathead rivers are the southernmost riverseast of the Continental Divide and the extreme uppermost in

the Columbia River system for bull trout, respectively(Fig. 5a). The nearest glacial refugium with known connec-tions to this region would have been the upper Columbia(Fig. 5b). Another possibility could have been the Missourirefugium (Haas 1988), but no bull trout presently exist in anyMissouri River drainages or museum collections (Haas andMcPhail 1991). The necessary postglacial watershed connec-tions and timing for such a recolonization from the upperColumbia refugium existed (St.-Onge 1972; Christiansen1979), and similar drainage affiliations are known in otherfish species (Lindsey and McPhail 1986; McPhail andLindsey 1986). The Jarbridge River in Nevada is an upstreamtributary to the Snake River, which then flows into the Co-lumbia River also at a more upstream location. The JarbridgeRiver may cluster with these systems, even though it was notdirectly impacted by Wisconsinan glaciation, and could alsobe the result of morphometric similarity and similar environ-ments. Previous glaciations had linkages within this overallregion, so its clustering could also reflect older common con-nections (Hubbs and Miller 1948; Smith 1978). All of thesemorphometric patterns here match those determined frommolecular genetic data (Bellerud et al. 1997; Spruell andAllendorf 1997; Taylor et al. 1999) and relate to those basedon other meristic characters taken from a more limited rangeand number of bull trout (Cavender 1997).

The Athabaska River, southern middle–upper ColumbiaRiver, eastern middle–extreme uppermost Fraser River, andthe Peace River east of its canyon barrier were likely post-glacially recolonized from the middle Columbia refugium(McPhail and Lindsey 1970, 1986). The Columbia refugearea was subdivided into distinct regions resulting from iso-lation by ice dams at various times in the Wisconsinan glaci-ation. The upper Columbia River was probably recolonizedfrom these central and southern middle sections of the Co-lumbia refugium as they are largely freely connected nowand were first postglacially connected by a series of large-scale downstream floods from the draining of huge glaciallakes (e.g., Bretz 1919). This postglacial flooding also oc-curred eastward through the Snake River (e.g., Malde 1968).

The northern middle Columbia River and eastern middleFraser River are clustered somewhat separately. This isintriguing because these drainage regions represent two setsof tributaries, the Similkameen–Okanagan rivers to theNicola–Coldwater rivers and the Arrow Lakes to the Thomp-son River, which formed the only large postglacial connec-tions between the Fraser and Columbia rivers (Mathews1944; Fulton 1969; Kershaw 1978). There is some similarand other fish species differentiation within the ColumbiaRiver (Hubbs and Miller 1948; Smith 1978). Bull trout arenow absent as sampled here and in museum collections fromat least the Canadian portions of the Similkameen River.

The Athabaska and extreme upper Fraser rivers likelyreceived bull trout as postglacial headwater crossovers.These rivers come very near each other, and postglacial ormodern flooding may have occurred here. The AthabaskaRiver and extreme upper Fraser River are also believed tohave been connected by a proglacial lake (Taylor 1960), butthe use of this connection by fish postglacially has been con-sidered unimportant (Lindsey and McPhail 1986). The east-ern Peace River may also have received bull trout throughpostglacial headwater transfer here as it, too, is very much in

© 2001 NRC Canada


this vicinity (Lindsey and McPhail 1986; McPhail andCarveth 1993).

The Peace River is naturally separated into distinct easternand western portions by a canyon barrier, now impoundedby a hydroelectric dam. Differentiation between fish types isnoticeable here for bull trout and for other fish species(Lindsey 1956; Lindsey and McPhail 1986; McPhail andCarveth 1993). The eastern Peace River is spread acrossboth minor dendrogram sections of our major middle Co-lumbia refugium division. More of its drainages do, how-ever, fall on the side of the northern middle Columbia andeastern middle Peace Rivers. Thus, it possibly is moreclosely related to these latter drainages and may instead havereceived bull trout through a postglacial headwater transferwith the Thompson River. The Thompson River comes veryclose to this convergence region but does not appear to havecontributed to the Athabaska or extreme upper Fraser rivers.Interplay of potential connections and dispersals, and theirtimings, successes, and failures, may have resulted in suchcomplex patterns (Lindsey 1956; Lindsey and McPhail 1986).As well, hypotheses could become more distinct or slightlychange with the addition of more samples, especially fromsuch regions less well represented. All these morphometricpatterns here match those determined from molecular ge-netic data (Bellerud et al. 1997; Spruell and Allendorf 1997;Taylor et al. 1999).

Coastal British Columbia, the lower–western, middle–upper Fraser River and the Peace River west of its formercanyon barrier were likely recolonized from an Chehalisrefugium, but they also may be partly from a lower Colum-bia refugium (McPhail and Lindsey 1986; McPhail andCarveth 1993). Because we could not get permission to col-lect and did not obtain specimens from the Olympic Penin-sula, we could not attempt to completely discern this pattern.The western Peace River is now a separate drainage but waspostglacially connected to the upper Fraser River (Tipper1971; McPhail and Lindsey 1986), and this recolonizationroute was likely used by several other fish species (Lindseyand McPhail 1986; McPhail and Carveth 1993).

Lower Columbia refugium bull trout do not appear to beanadromous, and bull trout in general do not show extensiveocean migration (Haas and McPhail 1991) but rather seem tomake smaller migrations, perhaps similar to those for coastalcutthroat trout (Oncorhynchus clarki clarki). Bull trout fromthese regions are an exception to this as there presently isunpublished evidence that they do at least regularly enter thesea and even move between these drainages. Those anadro-mous bull trout in Wisconsinan-glaciated parts of this regioncould be speculated to be from the Chehalis refugium,whereas nonanadromous bull trout could be from the lowerColumbia refugium.

The Olympic Peninsula region would thus be the likelyrefugium candidate for the downstream coastal parts of thisregion because it is closer, had the first postglacial access,and had freshwater postglacial connections (Armstrong1981; McPhail and Carveth 1993). The refugium is gener-ally referred to as the Chehalis. The lower Columbiarefugium bull trout are also clustered with the glaciated por-tions of this area but may have recolonized its more up-stream reaches. These upstream parts are above the highwater velocity barrier of the Fraser River Canyon, which

remains an upstream restriction with similar differentiationpatterns for other fish species (e.g., McPhail and Lindsey1986). Nonetheless, lower Columbia refugium fish speciescould also have recolonized some coastal areas of these re-gions as is suspected for other fish species (e.g., Reimersand Bond 1967).

Interestingly, Haas and McPhail (1991; colour photographsin Haas 1988) documented that some populations of lowerFraser River bull trout did not take on typical spawning mor-phology and coloration. This was considered a possible con-sequence of living in proximity with Dolly Varden(S. malma), but perhaps it could be because there are twobiogeographic types of bull trout in this region. Othersympatric bull trout and Dolly Varden situations discoveredsince then do have both char species taking on typicalspawning characteristics.

The Klamath and California river drainages are the westernregions here not Wisconsinan glaciated. They likely clusterhere and together because they were unglaciated and aregeographically proximal. This could indicate some similaritybetween them, but they also still group with the lower Co-lumbia River. Bull trout from the Columbia and Klamathrivers are genetically distinct based on allozyme data (Learyet al. 1993; Bellerud et al. 1997; Spruell and Allendorf1997). Distinctions based on other data are also the case forsome other fish species here (e.g., Hughes et al. 1987). Someof these relationships likely predate the last Pleistocene gla-ciation and its postglacial recolonization and relationshippatterns. The overall relationships between glaciated andthese much older unglaciated regions may be too difficult todecipher given the multiple refugia and unglaciated regionswith complex pre-Wisconsinan glaciation connections(Hubbs and Miller 1948; Miller 1959; Smith 1978). As well,our sample sizes for the California and Klamath riverdrainages are very small because these populations arerespectively extinct and most endangered (e.g., Leary et al.1993). Smaller scale or more extensive analyses of thesegroups might differentiate them further. Nonetheless, allthese morphometric patterns match those available frommolecular genetic data (Bellerud et al. 1997; Spruell andAllendorf 1997; Taylor et al. 1999) and relate to those basedon other meristic characteristics taken from a more limitedrange and number of bull trout (Cavender 1997).

The separate clustering of coastal British Columbia andthe spreading of the upper Fraser and western Peace riversacross this biogeographic group is at a low Euclidean dis-tance. This is also the case for the eastern Peace River in theprevious major right central dendrogram split correspondingto the upper Columbia refugium. These overall distinctionsmay simply not be as strong. For instance, the east–westsplit in the Peace River corresponding to its canyon was notalways completely definite during a minority of the jack-knife trials. Some populations on the immediate either sideof the canyon are sometimes clustered with the oppositegroup. An alternative explanation is that there may havebeen crossovers between drainages and introgressions be-tween glacial refugia types here (Foote et al. 1992; McPhailand Carveth 1993).

The northwestern river systems of the Liard, Nass, Skeena,Stikine, and Yukon rivers could have had an independentglacial origin from the Bering and (or) Nahanni refugia.

© 2001 NRC Canada


Foote et al. (1992) postulated largely on the basis ofallozyme data that lake whitefish (Coregonus clupeaformis)in the Yukon River and in this region of the Liard Riverwere recolonized from the Bering refugium. However, laketrout (S. namaycush) in this same area of the Liard Riverwere postulated on the basis of mtDNA to be from aNahanni refugium, with their absence from much of the Yu-kon River drainage also supporting this contention (Wilsonand Hebert 1998; also see Lindsey 1964). The Nahannirefugium is not accepted as definitive by all researchers. Thespecific bull trout sample site localites here were not exam-ined for either of these other species.

Because the Liard River bull trout cluster with those ofthe Nass and Skeena rivers, it is conceivable that these threerivers were recolonized from the Nahanni refugium and thatthe Yukon and Stikine rivers were recolonized from theBering. Complex glacial history (Ford 1976; Mathews 1980)and postglacial drainage connections (Templeman-Kluit1980; White et al. 1985; Rampton 1987) in this region couldsupport almost any hypotheses. More samples might resolvethis differentiation, or perhaps there is sufficient crossoverthat it would still be difficult (Lindsey and McPhail 1986;McPhail and Lindsey 1986; Foote et al. 1992). We believethat bull trout exist further north (McPhail 1961; Lindseyand McPhail 1986; Haas and McPhail 1991), and if they hadbeen collected or available, these results might have beenclarified.

This distinction for the northwestern river systems is theonly real difference to the biogeographic scenario deter-mined from molecular genetic data (Taylor et al. 1999), butit does still relate to that based on other meristic characterstaken from a more limited range and number of bull trout(Cavender 1997). The DNA results suggest that this regionwas recolonized from portions of the Columbia refugium.The separate clustering of these most northern groups couldconceivably in part be due to their particularly distant geo-graphic isolation. Alternatively, they may be indicating vari-ation or relationships that do not show up in the moleculargenetic data.

Ecological biogeographyThe hypothesized broad ecological biogeographic patterns

for bull trout are most easily interpreted by a three-grouprepresentation. The left group contains populations ofdrainages that were not Wisconsinan glaciated and arestrongly believed to all be nonanadromous. The remainingtwo large divisions represent Wisconsinan-glaciated regionsand are roughly split into western and eastern–northern seg-ments. This further split corresponds to an increase inanadromy, with many of the western populations recognizedas anadromous and all of the eastern–northern ones knownnot to be. The eastern–northern localities may still migratebut do not do so through the ocean.

The anadromous–nonanadromous grouping can even beseen within single river systems. The upper regions of theFraser and Skeena rivers are represented on the right non-anadromous side, whereas their lower reaches are on themiddle anadromous side. These lower reaches are closer tothe ocean and contain anadromous bull trout. Other coastalrivers like the Nass should not be assumed to have only

nonanadromous bull trout because anadromous samples maynot have been collected or available.

These ecological biogeographic results also correspondclosely with what would be expected for explanations of ourhistorical biogeographic hypotheses of particular glacialrefugia and postglacial recolonization routes. If any bulltrout populations were postglacially recolonized from thelower Columbia refugium, they do appear nonanadromous.Any postglacial recolonization of the lower Fraser River andsouthern coastal British Columbia would seem to have hadto be through freshwater routes or through freshwater flood-ing of inshore ocean regions. These Wisconsinan-glaciatedsouthern coastal localities are all represented here as anadro-mous, so it would appear more likely that they were post-glacially recolonized largely from the Chehalis refugiumwhere anadromous bull trout exist. Again, we unfortunatelydid not have samples from the Olympic Peninsula, butnearby, related and representative coastal Washington fishare clustered in this anadromy branch of the dendrogram.We also now know that bull trout in these regions can beanadromous. It is worth noting that anadromous bull troutonly appear to exist where their coastal ranges overlap withthat of the similar char species Dolly Varden (Haas andMcPhail 1991).

The middle Fraser River has bull trout samples that likelyare nonanadromous and these could have been postglaciallyrecolonized from the rest of the Columbia refugium. Thismight further confirm the aforementioned historical bio-geographic hypothesis that two bull trout types may haverecolonized this general area. The other cluster ofWisconsinan-glaciated drainages that are nonanadromousrepresent those historical biogeographic drainages hypothe-sized to be recolonized from the Nahanni–Bering and upperand middle Columbia refugia.

Summary of biogeographic analyses—combinedhistorical and ecological approaches

The lack of congruence and interaction between the tradi-tional ecological and historical approaches to biogeographycan be resolved using CCA. CCA can quantitatively parti-tion historical and ecological information from morpho-metric data and construct detailed biogeographic hypotheseslargely reconciling these two otherwise usually completelyconfounded features. Other advantages are that such analy-ses can be done for individual and highly variable speciesand that they can be based on multivariate data sets that tendto be more representative of them. The historical and ecolog-ical biogeographic interpretations from the CCA werestrengthened and clarified by awareness and understandingof each other and their interactions. The historical bio-geographic patterns were also largely congruent with molec-ular genetic studies (Bellerud et al. 1997; Spruell andAllendorf 1997; Taylor et al. 1999).

This success of the CCA was in spite of bull trout beingnotoriously variable and plastic and occurring in westernNorth America, which has a very complex geological his-tory. In these areas, vicariance patterns are young and poten-tially of different ages (e.g., Cracraft 1988), and dispersal isobviously important (Mayden 1988). As well, much of thevariation to be examined here will likely be within species,particularly fish which have populations most readily and

© 2001 NRC Canada


often completely isolated from each other (McPhail andCarveth 1993; Haas 2000). Good congruence betweenmorphometric and genetic data is also not necessarily ex-pected because morphology should have a larger environ-mental component and be more directly susceptible tonatural selection. Similar biogeographic analyses might beparticularly insightful for less well-funded and thus oftenless well-researched species that have decent museum col-lections from regions recently affected by phenomena suchas glaciation.

Such multivariate techniques are still thought to work beston large-scale trends (Wartenberg 1985b) and may not besuitable for all biogeographic analyses. Other multivariateprocedures may be better at smaller scales and particularlyat examining ecological patterns in more detail. CCA ap-pears to work best with large data sets that cover a species’distribution well. Our data suggest that isolated samples orpoorly sampled regions may sometimes be difficult to inter-pret and are somewhat more prone to misrepresentation.Some of our biogeographic hypotheses could be modified ifmore samples from specific regions were added to the analy-ses (Hughes et al. 1987), even though our jackknifing proce-dures suggest that our results based on these availablesamples are robust.

Conclusions for conservation biology and fisheriesmanagement

The interplay of historical and ecological features shouldbe considered in research and in management (e.g., Gilbert1980; Rahel 1986; Claytor et al. 1991) and conservation pro-grams (e.g., Brooks et al. 1992; Mayden 1992), particularlyfor organisms in regions in which much of the variability isstill within species (e.g., Haas 2000). History and ecologycan have complex interactive effects (e.g., Rosen 1988;Gould and Woodruff 1990), and an appreciation for, andconsideration of, both features is critical if their otherwiseconfounded information is to be reconciled. This is particu-larly true of species in regions composed of high levels ofwithin-species variability. If, unfortunately or realistically,only certain populations or regions of a species could beconserved, such integrated analyses might help prioritizewhich of them would be critically representative of the mosthistorical or ecological variability and even which may standthe best chances of success.

The variabilities evolved through specific historical andecological events can have different effects on susceptibilityto perturbation and extinction (Brooks et al. 1992; Gorman1992) and are both worthy of preservation. Organisms inregions recently recolonized are likely composed more ofwithin-species variability, particularly for fish and otheraquatic taxa. Such historical and ecological informationwould have to be deciphered and integrated, preferably inanalyses of a more complete regional fauna and (or) flora, todetermine how many and which areas would be representa-tive of as much of the evolution and ecology of particularpopulations, species, or their assemblages as is possible orpractical to protect.

For bull trout, there appears to be distinct historicalgroupings and the most significant ecological componentwithin them and their present-day distribution seems to bemigratory life history. This suggests at a bare minimum that

as many bull trout groups as possible representing theirlikely historical variabilities, and within those groups alltheir migratory life histories and relationships, must be fullyconserved to be successful and representative. The size, lo-cation, and environmental and biological statuses of specificbull trout groups might further prioritize and guide conser-vation measures. Fisheries management should also accountfor and maintain such differences, with a single broad proto-col likely being inappropriate.

Acknowledgements

The many people who helped with the original char sys-tematics and distribution study are extensively and gratefullyacknowledged in Haas and McPhail (1991). Additionally, wesincerely thank Dr. M. Blouw, L. Huato Soberanis, Dr. D.Markle, B.T.D. Muttley, F.P. Muttley, and particularly D.Atagi, G. Birch, M. Nevin-Haas, and Dr. E. Taylor for dis-cussions or encouragement. D. Atagi, J. Baxter, and Dr. J.Maze kindly reviewed the initial drafts. Two anonymous re-viewers and the journal editorial staff also contributed sub-stantially to this document. This research was supportedlargely by Natural Sciences and Engineering ResearchCouncil of Canada (NSERC) grant No. A6451 to JDM.GRH also received personal funding through a British Co-lumbia Science Council Graduate Research Engineering andTechnology (GREAT) award in cooperation with B.C. HydroEnvironmental Resources, which also provided some subse-quent funding.

References

Andrews, D.F., Gnanadesikan, R., and Warner, J.L. 1973. Methodsfor assessing multivariate normality. In Multivariate analysis.III. Proceedings of the Third International Symposium on Multi-variate Analysis. Edited by P.R. Krishnaiah. Academic Press,New York. pp. 95–116

Armstrong, J.E. 1981. Post-Vashon Wisconsin glaciation, FraserLowland, British Columbia. Bull. Geol. Surv. Can. 322: 1–34.

Ball, I.R. 1975. Nature and formation of biogeographical hypothe-ses. Syst. Zool. 24: 407–430.

Balon, E.K. 1984. Life histories of Arctic charrs: an epigenetic ex-planation of their invading ability and evolution. In Biology ofthe Arctic charr: Proceedings of the International Symposium onArctic charr. Edited by L. Johnson and B.L. Burns. University ofManitoba Press, Winnipeg, Man. pp. 109–141.

Bellerud, B.L., Gunkel, S., Hemmingsen, A.R., Buchanan, D.V., andHowell, P.J. 1997. Bull trout life history, genetics, habitat needs,and limiting factors in central and northeast Oregon. 1996 AnnualReport for Bonneville Power Administration (P.O. Box 3621,Portland, OR 97208-3621, U.S.A.), Project No. 95-54 and Con-tract No. 94BI34342.

Birks, H.J.B. 1987. Recent methodological developments in quan-titative descriptive biogeography. Ann. Zool. Fenn. 24: 165–178.

Bookstein, F.L., Chernoff, B., Elder, R., Humphries, J., Smith, G.,and Strauss, R. 1985. Morphometrics in evolutionary biology.Acad. Nat. Sci. Philadelphia, Spec. Publ. 15.

Bretz, H.L. 1919. The late Pleistocene submergence in the Colum-bia Valley of Oregon and Washington. J. Geol. 27: 489–505.

Brooks, D.R. 1985. Historical ecology: a new approach to studyingthe evolution of ecological associations. Ann. Mo. Bot. Gard.72: 660–680.

© 2001 NRC Canada


© 2001 NRC Canada


Brooks, D.R., Mayden, R.L., and McLennan, D.A. 1992. Phylog-eny and biodiversity: conserving our evolutionary legacy. TrendsEcol. Evol. 7: 55–59.

Cavender, T.M. 1997. Morphological distinction of bull trout from theMcCloud River system of northern California. In Friends of theBull Trout Conference Proceedings. Edited by W.C. Mackay, M.K.Brewin, and M. Monita. Bull Trout Task Force (Alberta), c/o TroutUnlimited Canada, Calgary, Alta., Canada. pp. 271–282.

Chernoff, B. 1982. Character variation among populations and theanalysis of biogeography. Am. Zool. 22: 425–439.

Christiansen, E.A. 1979. The Wisconsin deglaciation of southernSaskatchewan and adjacent areas. Can. J. Earth Sci. 16: 913–938.

Claytor, R.R., MacCrimmon, H.R., and Gots, B.L. 1991. Continentaland ecological variance components of European and NorthAmerican Atlantic salmon, Salmo salar phenotypes. Biol. J. Linn.Soc. 44: 203–229.

Cracraft, J. 1988. Deep-history biogeography: retrieving the historicalpattern of evolving continental biotas. Syst. Zool. 37: 221–236.

Dunham. J.B., and Rieman, B.E. 1999. Metapopulation structure ofbull trout: influences of physical, biotic, and geometrical land-scape characteristics. Ecol. Appl. 92: 642–655.

Endler, J.A. 1982a. Alternative hypotheses in biogeography: intro-duction and synopsis of the symposium. Am. Zool. 22: 349–354.

Endler, J.A. 1982b. Problems in distinguishing historical from eco-logical factors in biogeography. Am. Zool. 22: 441–452.

Falconer, D.S. 1981. Introduction to quantitative genetics. 2nd ed.Longman House, London, U.K.

Filliben, J.J. 1975. The probability plot correlation coefficient testfor normality. Technometrics, 17: 111–117.

Foote, C.J., Clayton, J.W., Lindsey, C.C., and Bodaly, R.A. 1992.Evolution of lake whitefish (Coregonus clupeaformis) in NorthAmerica during the Pleistocene: evidence for a Nahanni glacialrefuge race in the northern Cordillera region. Can. J. Fish. Aquat.Sci. 49: 760–768.

Ford, D.C. 1976. Evidence of multiple glaciations in South NahanniNational Park, Mackenzie Mountains, Northwest Territories. Can.J. Earth Sci. 13: 1433–1445.

Fulton, R.J. 1969. Glacial lake history, southern interior plateau,British Columbia. Geol. Surv. Can. Pap. No. 37-69.

Gilbert, C.R. 1980. Zoogeographic factors in relation to biologicalmonitoring of fish. In Biological monitoring of fish. Edited byC.H. Hocutt and J.R. Stauffer. Lexington Books, D.C. Heath andCo., Lexington, Mass. pp. 309–355.

Gittins, R. 1979. Ecological applications of canonical analysis. InMultivariate methods in ecological work. Edited by C.R. Raoand W.M. Stiteler. International Co-operative Publishing House,Fairland, Md. pp. 309–535.

Gorman, O.T. 1992. Evolutionary ecology and historical ecology:assembly, structure, and organization of stream fish communi-ties. In Systematics, historical ecology, and North Americanfreshwater fishes. Edited by R.L. Mayden. Stanford UniversityPress, Stanford, Calif. pp. 659–688.

Gould, S.J., and Woodruff, D.S. 1990. History as a cause of areaeffects: an illustration from Cerion on Great Inagua, Bahamas.Biol. J. Linn. Soc. 40: 67–98.

Gower, J.C., and Ross, G.J.S. 1969. Minimum spanning trees andsingle linkage cluster analysis. Appl. Stat. 18: 54–64.

Green, P.E. 1978. Analyzing multivariate data. Dryden Press,Hinsdale, Ill.

Haas, G.R. 1988. The systematics, zoogeography, and evolution ofDolly Varden and bull trout in British Columbia. M.Sc. thesis,University of British Columbia, Vancouver, B.C.

Haas, G.R. 2000. British Columbia’s freshwater fish, species, andaquatic ecosystems are more at risk and less protected. In Pro-

ceedings of a Conference on the Biology and Management ofSpecies and Habitats at Risk, University College of the Cariboo,Kamloops, B.C., Canada, February 15–19, 1999. Edited by L.Darling. Crown Publications, Victoria, B.C. pp. 561–576.

Haas, G.R., and McPhail, J.D. 1991. Systematics and distributionsof Dolly Varden (Salvelinus malma) and bull trout (Salvelinusconfluentus) in North America. Can. J. Fish. Aquat. Sci. 48:2191–2211.

Hubbs, C.L., and Miller, R.R. 1948. The zoological evidence. Cor-relation between fish distribution and hydrographic history inthe desert basins of western United States. In The Great Basin,with special emphasis on glacial and post-glacial times. Univ.Utah Bull. No. 38. pp. 1–191.

Hughes, R.M., Rexstad, E., and Bond, C.E. 1987. The relationship ofaquatic ecoregions, river basins and physiographic provinces to theichthyogeographic regions of Oregon. Copeia, 1987: 423–432.

Kershaw, A.C. 1978. The quaternary history of the Okanagan. In42nd Annual Report of the Okanagan Historical Society. Editedby D. Thompson. Wayside Press, Vernon, B.C. pp. 7–42.

Leary, R.F., Allendorf, F.W., and Forbes, S.H. 1993. Conservation ge-netics of bull trout in the Columbia and Klamath River drainages.Conserv. Biol. 7: 856–865.

Lindsey, C.C. 1956. Distribution and taxonomy of fishes in theMacKenzie drainage of British Columbia. J. Fish. Res. BoardCan. 13: 759–789.

Lindsey, C.C. 1964. Problems in zoogeography of the lake trout,Salvelinus namaycush. J. Fish. Res. Board Can. 21: 977–994.

Lindsey, C.C., and McPhail, J.D. 1986. Zoogeography of fishes ofthe Yukon and Mackenzie basins. In The zoogeography of NorthAmerican freshwater fishes. Edited by C.H. Hocutt and E.O.Wiley. John Wiley and Sons, New York. pp. 639–674.

MacArthur, R.H., and Wilson, E.O. 1967. The theory of islandbiogeography. Princeton University Press, Princeton, N.J.

Malde, H.E. 1968. The catastrophic late Pleistocene Bonnevilleflood in the Snake River Plain, Idaho. U.S. Geol. Surv. Prof.Pap. No. 596.

Mathews, W.H. 1944. Glacial lakes and ice-retreat in south-centralBritish Columbia. Trans. Royal Soc. Can., Sect. IV, Ser. 2, 38:39–57.

Mathews, W.H. 1980. Retreat of the last ice sheets in northeasternBritish Columbia and adjacent Alberta. Bull. Geol. Surv. Can.No. 331.

Mayden, R.L. 1988. Vicariance biogeography, parsimony, and evolu-tion in North American freshwater fishes. Syst. Zool. 37: 329–355.

Mayden, R.L. 1992. An emerging revolution in comparative biol-ogy and the evolution of North American freshwater fishes. InSystematics, historical ecology, and North American freshwaterfishes. Edited by R.L. Mayden. Stanford University Press, Stan-ford, Calif. pp. 864–890.

McPhail, J.D. 1961. A systematic study of the Salvelinus alpinuscomplex in North America. J. Fish. Res. Board Can. 18: 793–816.

McPhail, J.D., and Carveth, R. 1993. A foundation for conserva-tion: the nature and origin of the freshwater fish fauna of BritishColumbia. Queen’s Printer for British Columbia, Victoria, B.C.(Available from author (J.D. McPhail).)

McPhail, J.D., and Lindsey, C.C. 1970. Freshwater fishes of north-western Canada and Alaska. Bull. Fish. Res. Board Can. No. 173.

McPhail, J.D., and Lindsey, C.C. 1986. Zoogeography of the fresh-water fishes of Cascadia (the Columbia system and rivers northto the Stikine). In The zoogeography of North American fresh-water fishes. Edited by C.H. Hocutt and E.O. Wiley. J. Wileyand Sons, New York. pp. 615–637.

Miller, R.R. 1959. Origin and affinities of the freshwater fish fauna

of western North America. In Zoogeography. Edited by C.L.Hubbs. Am. Assoc. Adv. Sci. Publ. No. 51. pp 187–222.

Milner, A.M., Knudsen, E.E., Soiseth, C., Robertson, A.L., Schell,D., Phillips, I.T., and Magnusson, K. 2000. Colonization and de-velopment of stream communities across a 200-year gradient inGlacier Bay National Park, Alaska, U.S.A. Can. J. Fish. Aquat.Sci. 57: 2319–2335.

Oswood, M.W., Reynolds, J.B., Irons III, J.G., and Milner, A.M.2000. Distributions of freshwater fishes in ecoregions and hydro-regions of Alaska. J. N. Am. Benthol. Soc. 19: 405–418.

Pielou, E.C. 1984. The interpretation of ecological data. A primer onclassification and ordination. John Wiley and Sons, New York.

Pimentel, R.A. 1979. Morphometrics. The multivariate analysis ofbiological data. Kendall/Hunt Co., Dubuque, Iowa.

Rahel, F.J. 1986. Biogeographic influences on fish species compo-sition of northern Wisconsin lakes with applications for lakeacidification studies. Can. J. Fish. Aquat. Sci. 43: 124–134.

Rampton, V.N. 1987. Late Wisconsin deglaciation and Holoceneriver evolution near Fort Nelson, northeastern British Columbia.Can. J. Earth Sci. 24: 188–191.

Reimers, P.E., and Bond, C.E. 1967. Distribution of the fishes intributaries of the lower Columbia River. Copeia, 1967: 541–550.

Reyment, R.A., Blackith, R.E., and Campbell, N.A. 1984. Multi-variate morphometrics. 2nd ed. Academic Press, London, U.K.

Rosen, D.R. 1988. Biogeographic patterns: a perceptual overview.In Analytical biogeography. An integrated approach to the studyof animal and plant distributions. Edited by A.A. Myers and P.S.Giller. Chapman and Hall, London, U.K. pp. 23–55.

Savvaitova, K.A. 1980. Taxonomy and biogeography of charrs in thePalearctic. In Charrs: salmonid fishes of the genus Salvelinus.Edited by E.K. Balon. Dr. W. Junk Publishers, The Hague, TheNetherlands. pp. 281–294.

Shea, B.T. 1985. Bivariate and multivariate growth allometry: sta-tistical and biological considerations. J. Zool. 206: 367–390.

Smith, G.R. 1978. Biogeography of intermountain fishes. InIntermountain biogeography, a symposium. Edited by K.T. Harperand J.L. Reveal. Great Basin Nat. Mem. No. 2. pp. 17–42.

Snyder, J.P. 1983. Map projections used by the U.S. GeologicalSurvey. 2nd ed. Bull. U.S. Geol. Surv. No. 1532.

Spruell, P., and Allendorf, F.W. 1997. Nuclear analysis of Oregonbull trout. Final report to Oregon Department of Fish and Wild-life, Wild Trout and Salmon Genetics Lab. Rep. No. 97/5. Uni-versity of Montana, Missoula, Mont. (Available from P. Spruell.)

Spruell, P., Rieman, B.E., Knudsen, K.L., Utter, F.M., and Allendorf,F.W. 1999. Genetic population structure within streams: micro-satellite analysis of bull trout populations. Ecol. Freshwater Fish,8: 114–121.

St.-Onge, D.A. 1972. Sequence of glacial lakes in north-central Al-berta. Bull. Geol. Surv. Can. No. 213.

Tabachnik, B.G., and Fidell, L.S. 1989. Using multivariate statis-tics. 2nd ed. Harper and Row Publishers, New York.

Taylor, D.W. 1960. Distribution of the freshwater clam, Pisidiumultramontanum: a zoogeographic inquiry. Am. J. Sci. 258A:325–334.

Taylor, E.B., Pollard, S., and Louie, D. 1999. Mitochondrial DNAvariation in bull trout (Salvelinus confluentus) from northwest-ern North America: implications for zoogeography and conser-vation. Mol. Ecol. 8: 1155–1170.

Templeman-Kluit, D. 1980. Evolution of physiography and drain-age in southern Yukon. Can. J. Earth Sci. 17: 1189–1203.

Thorpe, R.S. 1976. Biometric analysis of geographic variation andracial affinities. Biol. Rev. 51: 407–452.

Tipper, H.W. 1971. Glacial geomorphology and Pleistocene historyof central British Columbia. Bull. Geol. Surv. Can. 196: 1–89.

Waples, R.S. 1995. Evolutionary significant units and the conserva-tion of biological diversity under the endangered species act.Am. Fish. Soc. Symp. 17: 8–27.

Wartenberg, D.E. 1985a. Canonical correlation analysis: a methodfor describing geographic patterns. Syst. Zool. 34: 259–279.

Wartenberg, D.E. 1985b. Multivariate spatial autocorrelation: amethod for exploratory geographical analysis. Geog. Anal. 17:263–283.

White, J.M., Mathewes, R.W., and Mathewes, W.H. 1985. LatePleistocene chronology and environment of the “ice-free corri-dor” of northwestern Alberta. Quat. Res. 24: 173–186.

Wiley, E.O. 1988. Vicariance biogeography. Ann. Rev. Ecol. Syst.19: 513–542.

Wilson, C.C., and Hebert, P.D.N. 1998. Phylogeography and post-glacial dispersal of lake trout (Salvelinus namaycush) in NorthAmerica. Can. J. Fish. Aquat. Sci. 55: 1010–1024.

Appendix

Sample site (and museum storage) localities and fishnumbers

We kept the same sample site locality numbers in the Ap-pendix as in Haas and McPhail (1991) for consistency andease of interpretation and comparison. One small changewas made as to where samples were placed in overall drain-age basins in Haas and McPhail (1991). Localities 163 and164 are actually in the middle Columbia River drainage, notin the lower Columbia River drainage as we had previouslyand erroneously indicated. This mistake had no impact onthe initial systematic study of Haas and McPhail (1991) be-cause biogeography was not analyzed there.

Museum abbreviationsCAS, California Academy of ScienceOSU, Oregon State UniversityUA, University of AlbertaUBC, University of British ColumbiaUW, University of Washington

California drainages (relict)150. McCloud River (CAS, 4 fish in 4 samples)151. Mount Shasta Hatchery (CAS, 1 fish)

Nevada drainage152. Jarbridge River (CAS, 5 fish)

Klamath River drainages—Oregon153. Boulder–Brownsworth–Demming–Leonard creeks (OSU,

26 fish in 6 samples)154. Cherry Creek (OSU, 2 fish in 2 samples)155. Sun Creek (OSU, 11 fish in 3 samples)156. Long Creek (OSU, 12 fish in 2 samples)

Lower Columbia River drainages—Oregon–Washington157. Trapper Creek (OSU, 5 fish)158. Anderson Creek (OSU, 6 fish)159. McKenzie River (OSU, 5 fish)160. Metolius River (OSU, 2 fish)161. Jefferson Creek (OSU, 2 fish)162. Candle–Jack–Roaring creeks (OSU, 26 fish)165. Pine Creek (OSU, 4 fish)166. White Salmon River (UW, 1 fish)

© 2001 NRC Canada


Middle Columbia River drainages—Oregon–Washington163. Big Creek (OSU, 4 fish)164. Malheur River (OSU, 2 fish)167. Lick Creek (OSU, 1 fish)168. Touchet River (UW, 1 fish)169. Rock Island Dam, Columbia River (CAS, 1 fish)170. Wenatchee, Columbia River (CAS, 2 fish)171. Entiat Wier, Columbia River (CAS, 1 fish)172. Methow River (UW, 2 fish)173. Stehekin River (OSU, 2 fish)

Coastal drainages—Washington174. Golden Gardens, King County (UW, 1 fish)175. Baker River (UW, 1 fish)

Lower Fraser River drainages1. Pitt Lake (UBC, 2 fish)

176. Spanish Banks (UBC, – 1 fish)177. Pitt River mouth (UBC, 2 fish)178. Gold Creek (UBC, 1 fish)179. Chehalis Lake (UBC, 9 fish)180. Vedder River (UBC, 7 fish in 4 samples)181. Foley Creek (UBC, 2 fish)182. Foley Lake (UBC, 2 fish)183. Coquihalla River (UBC, 7 fish)

Middle Fraser River drainages184. Nicola River (UBC, 1 fish)185. Nicola Lake (UBC, 1 fish)186. Coldwater River (UBC, 2 fish)187. Birkenhead River (CAS, 3 fish)188. Lake Creek (CAS, 1 fish)189. Seton Lake (UBC, 3 fish)190. Anderson Creek (UBC, 1 fish)191. Fraser River (UBC, 6 fish)192. Kamloops Lake (UBC, 2 fish)193. Yard Creek (UBC, 1 fish)194. Anderson Lake (UBC, 1 fish)

Coastal drainages—British Columbia195. Alice Lake (UBC, 1 fish)196. Bluff Lake (UBC, 4 fish)197. One Eye Lake (UBC, 2 fish)

Upper Columbia River drainages—British Columbia–Montana

198. Salmo River (UBC, 13 fish)199. Kootenay Lake (UBC, 1 fish)200. Powder Creek (UBC, 1 fish)201. Lardeau River (UBC, 2 fish)202. Duncan River (UBC, 3 fish in 2 samples)203. Trout Lake (UBC, 1 fish)204. St. Leon Creek (UBC, 10 fish in 4 samples)205. Mackenzie Creek (UBC, 15 fish in 3 samples)206. Greely Creek (UBC, 6 fish)207. Illicillewaet River (UBC, 2 fish)208. Woolsey (Silver) Creek (UBC, 10 fish)209. South Pass Creek (UBC, 1 fish)210. Flat Creek (UBC, 16 fish)211. Columbia River (UBC, 2 fish)212. Blaeberry River (UBC, 6 fish)

213. Kickinghorse River (UBC, 2 fish)214. Columbia Lake (UBC, 1 fish)215. Whitetail Lake (UBC, 1 fish)216. St. Mary Lake (UBC, 3 fish)217. Moyie Lake (UBC, 12 fish in 2 samples)218. Gold Creek (UBC, 1 fish)219. Lizard Creek (UBC, 1 fish)220. Flathead River tributary (UBC, 10 fish)221. Shepp Creek (UBC, 8 fish)222. Sage Creek (UBC, 15 fish)223. Howell Creek (UBC, 3 fish)224. Flathead River (UBC, 3 fish)225. Inez Lake (OSU, 1 fish)226. Seeley Lake (OSU, 2 fish)

South Saskatchewan River drainages227. Irabel (UW, 3 fish)228. Kennedy Creek (UW, 1 fish)229. Lower Kananaskis Lake (UA, 1 fish)230. Smith-Dorrien Creek (UA, 8 fish)231. Barrier Reservoir (UA, 1 fish)232. Prairie Creek (UA, 1 fish)233. Sheep River (UA, 1 fish)

North Saskatchewan River drainages234. Upper North Saskatchewan River (UA, 9 fish in 2

samples)235. North Saskatchewan River (UA, 2 fish)236. Siffleur River (UA, 2 fish)237. Abraham Lake (UA, 19 fish in 5 samples)238. Cline River / Reservoir (UA, 10 fish in 3 samples)239. North Saskatchewan River (UA, 7 fish in 4 samples)240. Malma Creek (UA, 1 fish)241. Whiterabbit Creek (UA, 16 fish in 3 samples)242. Haven Creek (UA, 5 fish)243. Tershishner Creek (UA, 2 fish in 2 samples)244. North Saskatchewan River (UA, 2 fish)245. Clearwater River (UA, 2 fish in 2 samples)246. Seven Mile Creek (UA, 2 fish)247. Elk Creek (UA, 2 fish)248. Chungo Creek (UA, 1 fish)249. Un-named Creek (UA, 1 fish)250. Brown Creek (UA, 1 fish)251. Southesk River (UA, 2 fish)

Athabaska River drainages252. McLeod River (UA, 1 fish)253. Deerlick / Eunice creeks (U,– 7 fish)254. Cold Creek (UA, 4 fish)255. Mason Creek (UA, 4 fish)256. Rock Creek (UA, 11 fish)257. Gorge Creek (UA, 1 fish)258. Chickadee Creek (UA, 1 fish)

Upper Fraser River drainages2. Middle River (UBC, 8 fish in 2 samples)3. Tahtsa River (UBC, 1 fish)

259. Quesnel River (UBC, 3 fish)260. Quesnel River (UBC, 1 fish)261. Swift River (UBC, 2 fish)262. Cottonwood River (UBC, 8 fish in 3 samples)

© 2001 NRC Canada


© 2001 NRC Canada


263. Hixon Creek (UBC, 6 fish)264. Stone Creek (UBC, 4 fish)265. Takla Lake (UBC, 1 fish)266. Wright Creek (UBC, 3 fish in 2 samples)267. Aleza Lake (UBC, 3 fish)268. Hansard Lake (UBC, 2 fish)269. Willow River (UBC, 1 fish)270. Bowron River (UBC, 2 fish)271. King Creek (UBC, 19 fish)272. Robson River (UBC, 1 fish)273. Yellowhead Lake (UBC, 1 fish)274. Swift Creek (UA, 2 fish)275. Albreda Camp Creek (UA, 4 fish; UBC, 2 fish)

Skeena River drainages4. Shames River (UBC, 13 fish in 2 samples)5. Silburn Creek (UBC, 6 fish)6. Seeley Lake (UBC, 9 fish)7. Kispiox River (UBC, 1 fish)

276. Morice Lake (UBC, 8 fish in 4 samples)277. Bulkley River (UBC, 1 fish)278. Dennis Lake (UBC, 5 fish)279. Kathlyn Creek (UBC, 3 fish)280. Suskwa River (UBC, 2 fish)281. Bulkley River (UBC, 2 fish)282. Kitsumkalum Lake (UBC, 2 fish in 2 samples)

Nass River drainages8. Meziadin Lake (UBC, 5 fish in 2 samples)9. Surprise Creek (UBC, 6 fish)

283. Nass River (UBC, 1 fish)284. Nass River (UBC, 3 fish)285. Bowser Lake (UBC, 8 fish)

Stikine River drainage10. Winter Creek (UBC, 4 fish)

Peace River drainages286. Lower Pierre Greys Lake (UA, 3 fish)287. Rocky Creek (UA, 4 fish)288. Little Smoky River (UA, 3 fish)289. Mast Creek (UBC, 3 fish)290. Murray River (UBC, 2 fish)291. Stoddart Creek, Charlie Lake (UBC, 1 fish)292. Hook Lake (UBC, 5 fish)293. Peace River (UBC, 1 fish)294. Avalanche River (UBC, 3 fish)295. Misinchinka Swamp (UBC, 1 fish)296. Tacheeda Lakes (UBC, 5 fish in 2 samples)297. Parsnip River (UBC, 1 fish)298. Sylvester Creek (UBC, 2 fish in 2 samples)299. Germanson Lake (UBC, 1 fish)300. Diver Lake (UBC, 1 fish)301. Chowade River (UBC, 1 fish)302. Toodoggone Lake (UBC, 1 fish)303. Chesterfield Lake (UBC, 1 fish)

Liard River drainages304. Manson Creek (UBC, 1 fish)305. Bass Creek (UBC, 2 fish)306. Simmons Lake (UBC, 1 fish)307. Bus Lake (UBC, 8 fish)308. Prairie Creek, Northwest Territories, Canada (UBC,

5 fish)

Yukon River drainages309. Swan Lake (UBC, 2 fish)310. Gladys Lake (UBC, 9 fish in 2 samples)

The post-Wisconsinan glacial biogeography of bull trout ... › rm › boise › publications...rately on the truss and meristic characters (e.g., Thorpe 1976) sug-gest that there

Documents