|
<br />242
<br />
<br />MICHAEL E. DOUGLAS ET AL.
<br />
<br />direct and indirect tectonic processes (e.g., formation of ba-
<br />sin-and-range topography and other drainage-disrupting
<br />events), the result being the now-existing, mosaic pattern of
<br />variation.
<br />Vicariance biogeography has largely been applied in a
<br />global perspective, because as spatial scale increases so does
<br />the probability that resulting biological patterns involve
<br />groups of phylogenetic ally related forms (Brooks 1988).
<br />Hence, to achieve interpretable patterns of an historical na-
<br />ture, broad questions such as "How are continental biotas
<br />assembled over time?" (Craycraft 1988) must be employed
<br />(see also Bermingham et al. 1992). Rosen (1974, 1979) pro-
<br />vided excellent examples of vicariance biogeography as re-
<br />lates to freshwater fish faunas.
<br />Rosen (1985) also argued for a particular approach when
<br />he stated "what is needed is a precise means of specifying
<br />how a given biohistory is explicitly tied to a particular geo-
<br />history." While arguing for cladistic methodology, Rosen
<br />(1985, p. 637) also recognized (in our perspective) a statis-
<br />tical approach (fostered herein), by stating, "it is this inde-
<br />pendence of biological from geological data that makes the
<br />comparison of the two so interesting because it is hard to
<br />imagine how congruence between the two could be the result
<br />of anything but a causal history in which geology acts as the
<br />independent variable providing opportunities for change in
<br />the dependent biological world. The comparison becomes
<br />especially interesting if there is a congruence among geo-
<br />histories based on different approaches to the geographic
<br />problem, and if there is a congruence among cladistic rela-
<br />tions of different taxa with regard to the same geographic
<br />areas. "
<br />Vicariant studies of a regional or local nature are few, the
<br />reasons for which are several. Numbers of species decrease
<br />with reduced area, which make patterns difficult to discern
<br />(Brooks 1988). Also, species patterns at the local or even
<br />regional level are often confounded by ecologic phenomena,
<br />such that historic patterns are often masked (Chernoff 1982;
<br />Endler 1982). Finally, a general lack of geologic information
<br />at the regional or local level makes vicariant events more
<br />difficult to interpret. Notable exceptions are Smith's (1983)
<br />association between patterns of glaciation and distributions
<br />of freshwater fishes in New York State. Also noteworthy is
<br />the treatment by Schneider et al. (1998) of patterns in genetic
<br />variation of frogs and lizards driven by historic climatic
<br />change in Australian rainforests.
<br />
<br />GEOGRAPHIC REPRESENT A nON OF MODELS
<br />
<br />The first matrix we compiled was for stream size (SIZ),
<br />ranked by mean volume of discharge over periods of pub-
<br />lished records (Brown et al. 1979; U.S. Geological Survey
<br />various dates), which varied from 25 to 100 plus years. The
<br />assumption here is that discharge relates directly to watershed
<br />size. Four categories were: 0 = 0.0; 1 = > 0.0-0.28; 2 =
<br />> 0.28-1.42; and 3 = > 1.42 m3/sec.
<br />The second comparative matrix consiste,d of elevations
<br />above present MSL (ELE; to the nearest 30 m) of sampling
<br />sites. This matrix is interpretable as a surrogate for thermal
<br />conditions, with cooler water prevailing at higher altitude.
<br />Possible latitudinal effects on temperature were ignored due
<br />
<br />to regional altitudinal ranges (< 350 to > 2000 m), which
<br />tend to mask latitudinal effects over the relatively small geo-
<br />graphic area examined.
<br />Third was a barrier (BAR) matrix created to represent the
<br />degree to which each sampling site on a tributary was isolated
<br />from its receiving river, with isolation being a surrogate for
<br />the probability of recent interspecific hybridization. Although
<br />essentially all major streams in the region are now modified,
<br />ephemeral, or fragmented by human intervention, all were
<br />historically confluent, at least seasonally, and most were pe-
<br />rennial. Thus, presence of a physical barrier would preclude
<br />hybridization, whereas absence of a barrier would allow po-
<br />tential contact and miscegenation. Based on field studies, a
<br />collecting site's degree of isolation was estimated as one of
<br />three states: 0, no physical barrier between a site and a main-
<br />stream (high probability for contact and genetic exchange);
<br />1, partial isolation, in which fish at a site were isolated by
<br />seasonally dry reaches, precipitous gradients, or a combi-
<br />nation of both (intermediate probability of contact); and 2,
<br />absolute isolation (low probability of hybridization) in which
<br />fish either occurred above barrier waterfalls or were within
<br />the Santa Cruz Basin, which historically failed to reach the
<br />Gila River mainstream in a distinct channel, even during flood
<br />(Hendrickson and Minckley 1985).
<br />Finally, we derived paleodrainage matrices by superim-
<br />posing sampling sites over sketch maps depicting drainage
<br />relations inferred from ancient basins and basin-fill deposits
<br />in what now is Arizona during different geologic epochs (Fig.
<br />2; Nations et al. 1982). Phenotypic diversity at existing sites
<br />could in this way be contrasted with approximate hydro-
<br />graphic relations of the same sites in earlier times. The tested
<br />hypothesis is that phenotypically similar populations now
<br />spatially disjunct were sympatric in the past. Six matrices
<br />were evaluated, each representing all or part of an epoch:
<br />Oligocene (OLG); Early Miocene (EMI), mid-Miocene
<br />(MMI); Pliocene (PLI); Pleistocene (PLE); and Recent (REC)
<br />subbasins. Sites were scored as follows, depending upon the
<br />subbasin in which they were located during a particular ep-
<br />och: 0, Santa Cruz; 1, upper San Pedro; 2, lower San Pedro;
<br />3, Gila; 4, Verde; 5, Salt; and 6, Agua Fria. All were then
<br />converted into matrices of taxonomic distance and each con-
<br />trasted against the phenotypic shape matrix using the Mantel
<br />test.
<br />
<br />\
<br />
<br />RESULTS AND DISCUSSION
<br />
<br />Matrix Correlation Analysis
<br />
<br />If the hypothesis for Model I (ecotypy or ecophenotypy-
<br />morphological response to local environments) is supported,
<br />shape (H) should be significantly and positively associated
<br />with matrices of either elevation (ELE), stream size (SIZ),
<br />or both. To satisfy Model II (hybridization or intergrada-
<br />tion-relative to isolation), H should be negatively associated
<br />with BAR (i.e., association among phenotypes would be sig-
<br />nificantly related to a high degree of connectedness among
<br />habitats). If Model III (vicariance-response to historic
<br />drainages or drainage patterns) is applicable, phenotypic data
<br />from existing sites should be significantly associated with
<br />one or more matrices representing hydrographic relations de-
<br />
|