2189 The post-Wisconsinan glacial biogeography of bull trout (Salvelinus confluentus): a multivariate morphometric approach for conservation biology and management Gordon R. Haas and J.D. McPhail Abstract: Canonical correlation analysis (CCA) can quantitatively partition historical and ecological information from morphometric data where these features are otherwise confounded. CCA is applied to sample site locality morphometric data and corresponding sample site locality coordinate data for bull trout. Two vectors result. The first accounts for the maximum morphometric variation correlated to geographic information specified by the locality coordinates. The second represents the remaining less correlated variation. For biogeography, the first vector generates historical hypotheses for Pleistocene glacial refugia and for post-Wisconsinan glacial recolonization patterns and phylogeographic relationships. The second vector infers hypotheses for broad ecological patterns. The historical biogeographic patterns for bull trout suggest recolonization from either two or three glacial refugia and emphasize within-species biodiversity in western North America. These patterns from the Chehalis and Columbia refugia are largely concordant with other analyses based on molecular genetics. The morphometric analysis also suggests the additional possibility of a Nahanni and (or) Bering refugium. The ecological patterns suggest the importance and extent of 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 et écologiques des données morphométriques, lorsque celles-ci sont entremêlées. Une analyse CCA menée sur des données morphométriques de l’Omble à tête plate reliées à des sites d’échantillonnage ainsi que sur les données du milieu correspondant 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 illustre le reste de la variation moins corrélée. En ce qui a trait à la biogéographie, le premier vecteur génère des hypothèses historiques sur les refuges glaciaires pendant le Pléistocène, sur les voies de recolonisation après les glaciations du Wisconsinien 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 de l’Amérique du Nord. Ces structures associées aux refuges de Chehalis et de Columbia s’accordent en grande partie avec les résultats d’analyses basées sur la génétique moléculaire. L’analyse morphométrique laisse croire à la possibilité d’un refuge additionnel dans le Nahanni et (ou) la Béringie. Les structures écologiques démontrent l’importance et l’étendue de la 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 Species Act in the United States (U.S.) (e.g., fish, see Waples 1995). Biogeography is seemingly split into two Biogeography is the study of the distribution of organisms approaches with different aims and time scales (e.g., Ball and their variability in space and time. This discipline is 1975; Endler 1982a; Birks 1987). The first, the ecological often used to discern groupings of animals for conservation school, studies the dispersion of organisms and the mecha- and management purposes such as within the Endangered nisms and environmental interactions that maintain or Received June 7, 2000. Accepted August 27, 2001. Published on the NRC Research Press Web site at http://cjfas.nrc.ca on November 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. Can. J. Fish. Aquat. Sci. 58: 2189–2203 (2001) DOI: 10.1139/cjfas-58-11-2189 © 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 collections were not permitted and samples were unavailable for our original research, and thus were also not analyzed for this study. For sample site locations, see corresponding numbers in the Appendix. change this dispersion (e.g., MacArthur and Wilson 1967). geographic patterns of single species (Brooks 1985; Cracraft This research is usually done at the population or commu- 1988; Gorman 1992). nity level and sometimes involves direct experimentation. The approaches of the ecological and historical schools to The second is the historical school that studies spatial and biogeography are thus generally separate and confounded. A temporal distribution patterns (e.g., Brooks 1985; Wiley complete biogeographic analysis for conservation or man- 1988). This work is conducted at the taxonomic level and at- agement purposes really should contend with the integration tempts to explain distributions and interactions using past of both ecological and historical information (e.g., Endler events. Direct experimentation is thus not generally possible 1982a, 1982b; Mayden 1992). A combined procedure should and explanations are usually inferential (Gould and Wood- discern which data aspects are related to these categories and ruff 1990). what it says about them singly and in relation to each other The ecological school focuses on local populations in (Cracraft 1988). This lack of congruence can result in diffi- which diversity often has developed within a short evolution- culty or bias in the rigorous analyses of unconfounded com- ary time scale. These researchers are usually not interested plete biogeographic patterns, especially for single species. in broad biogeographic or species-specific patterns and con- Canonical correlation analysis (CCA) could be used in a sequently often ignore the potential influence of historical combined ecological and historical approach within bio- events and phylogeny. The historical school finds small- geography. CCA can quantitatively separate confounding in- scale ecological differentiation problematic. Their tendency formation, uses multivariate data sets, and operates at all is to work at a higher taxonomic level and look for general levels. For our biogeographic analyses, CCA essentially biogeographic patterns. Historical analyses usually do not quantifies the multivariate relationship between a set of deal with subspecific or localized variation or with bio- morphometric and corresponding locality coordinate matri- © 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). For sample site locations, see corresponding numbers in the Appendix. ces. The maximum morphometric variation correlated to the 1986), with char (Salvelinus spp.) being well recognized for geographic information specified by the locality coordinates their rapid postglacial recolonization (Balon 1984; Milner et generates hypotheses for historical biogeographic patterns. al. 2000; Oswood et al. 2000). Bull trout often exhibit strong The remaining variation not as strongly correlated or even migrations and their life history is usually summarized into uncorrelated to geography is used to infer hypotheses of four migratory types: resident, fluvial, adfluvial, and anadro- broad ecological biogeographic patterns. These historical mous or at least sea-run. The genetic variability of bull trout and ecological features could then be integrated and exam- is usually low within populations and often marked between ined in relation to each other for a complete and detailed them (e.g., Leary et al. 1993; Taylor et al. 1999), and a biogeographic analysis. metapopulation model has consistently been invoked to To demonstrate, we use CCA to analyze the biogeography account for this, at least in watersheds on a locally intercon- of bull trout (Salvelinus confluentus) throughout its broad nected scale (e.g., Dunham and Rieman 1999; Spruell et al. range (Haas and McPhail 1991). Fish are ideal for bio- 1999). These morphological and genetic patterns of variabil- geographic study because they are restricted to living in wa- ity for bull trout, combined with the biogeographic complex- ter and their distribution patterns must thus reflect specific ity of western North America, should appropriately test the discernible aquatic connections. Bull trout are also long ability of CCA to partition historical and ecological variation. renowned for their plasticity, which has previously con- Finally, other biogeographic analyses