Why does biological taxonomy matter

Morphological estimates of diversity

Estimates of the number of extant species of birds varied greatly during the first half of the 20th century [6], but stabilized with the check-list begun by Peters [7], edited by Peters, Mayr, Paynter and others through 16 volumes, and completed in 1987 [8]. These volumes treated many taxa as subspecies that had formerly been ranked as species, reflecting the philosophy of the proponents of the polytypic biological species concept prevalent at the time [6]. The completion of the series led to reference lists of the birds of the world [9] that were used in managing museum collections and, with a few updates, a list including 9159 species was adopted by the American Ornithologists' Union (AOU) Committee on Collections [10].

At about this time, molecular surveys of geographic variation began to appear and provide novel data on gene flow and genetic divergence. As a consequence, later compendia, for example, Sibley and Monroe [11], the 16 volumes of del Hoyo et al. [12], and most recently the Howard and Moore checklist[13,14], represent an eclectic mixture [15,16] of traditional morphological taxonomy combined with various molecular assessments of genetic differentiation [16]. These latter studies are still far from complete and, as a result, there currently exists no homogeneous appraisal of avian species taxa in the post-phenotypic era [17].

Using a uniform random number generator, we produced a sample of 200 species from the worldwide list of Wood & Schnell [10] of 9159 biological species of birds (S1 Table). This list reflects species limits as deduced mostly through morphological assessments. We deemed this list appropriate for testing morphological species limits. Although a small number of new species has been discovered since 1986, our purpose was to obtain a random sample such that the future number of newly described species would not measurably influence the study design (see below). In estimating the number of phylogenetic species (see S1 Appendix) within each biological species, we examined all specimens present for each species in the large specimen holdings of the American Museum of Natural History (New York), and in some instances, the collections of the Museum of Natural Science, Louisiana State University (Baton Rouge), and the Field Museum of Natural History (Chicago). Series of specimens representing subspecies or different geographic areas were examined for diagnosable differences (although specific catalog numbers were not recorded, catalog numbers of specimens examined can be obtained from the respective institutions for the species studied). The primary literature on geographic variation, as well as major faunistic works treating regions or relevant taxa, were also consulted. We used qualitative differences in plumage pattern, color, and morphology, including size and shape, as evidence of diagnosably distinct populations (i.e., phylogenetic species), whereas we ignored quantitative differences of size and shade of color. Biological species composed entirely of individuals without discernable qualitative differences within sex or age classes were considered a single phylogenetic species. Species with two or more distinctive morphologies within or among age or sex classes in the same localities were also considered to represent single polymorphic species. Biological species were hypothesized to contain multiple phylogenetic species if, within a single stratum of age or sex, populations showed at least one qualitative character difference that was either geographically disjunct (allopatric) or was continuous only by virtue of a narrow area of character change (gradient or cline) compared to the total geographic ranges of the different forms.

Three authors (GFB, JC, RMZ) individually investigated one third of the species in the random sample, and estimated the number of phylogenetic species included in each of the biological species. We computed the distribution of phylogenetic taxa per biological species for each of the three authors as well as the overall distribution. We used a bootstrap procedure to obtain confidence intervals for the average; this was done by resampling the distribution with replacement 1000 times. We also compared the distributions of the three authors using two sample Kolmogorov-Smirnov tests. To compare the results with generally accepted subspecific taxonomy of birds, we correlated our estimates of the numbers of phylogenetic species with the numbers of described subspecies in the same biological species-level taxon using a standard reference [8].

To estimate the effect of newly discovered species, recent molecular data, and changing taxonomic standards on our estimate of the number of phylogenetic species, we examined the classification of our sample of 200 species in the recent Howard and Moore checklists [13,14]. These lists comprise 10021 species that are partly based on phenotypic and partly based on genetic results. We determined whether each of the 200 species was treated identically to the AOU list, was lumped, split, or was otherwise altered in taxonomic treatment. In this way we obtained a ratio of phylogenetic species per biological species for the Howard and Moore 2013 and 2014 estimates of 10021 taxa.

Genetic estimates of diversity

For a more focused assessment of the effect of molecular studies on estimates of avian diversity, we surveyed the literature on phylogeographic studies, looking for variation and differentiation in rapidly evolving mitochondrial DNA (mtDNA) genes for species that had been well sampled from throughout their geographic range. In all we identified 437 biological species that were included in such studies and summarized the number of reciprocally monophyletic, geographically contiguous groupings within each (S2 Table). For analytical purposes we assume that these reciprocally monophyletic populations or groups of populations provided a proxy for phylogenetic species. This meta-analysis provides an independent genetic assessment of how many independent evolutionary lineages exist per biological species [18–20], and complements the morphological study.

We note two caveats. Species are generally not sampled randomly for molecular assessment of geographic structure (phylogeography). In general, investigators tend to choose species that they suspect a priori exhibit interesting patterns of geographic variation [21]. Thus, there may be a bias towards a higher phylogenetic to traditional species ratio for molecular studies. Secondly, molecular studies have shown that tropical species tend to show greater levels of differentiation than those of temperate species [22,23], and therefore estimates of the total number of taxa/species worldwide will be biased if the distribution of sampled species does not coincide with the geographic distribution of species diversity. We investigated this by plotting the distribution of phylogroups, generally recognized as reciprocally monophyletic populations on a mtDNA gene tree, as a function of latitude. To explore the relationship between the geographic distribution of biological species diversity, our sampling of biological species and the estimate of phylogenetic diversity, we extracted the average number of biological species in 10-degree latitude blocks from Figure 2 of Blackburn and Gaston [24]. We note there could be other biases, such as greater emphasis on New World taxa. To demonstrate the importance of including phylogroup taxa in evolutionary, ecological and biogeographical studies, we revisited the meta-analysis of Weir and Schluter [25]. They fit birth-death models to plots of molecular divergence dates vs. midpoint latitude, for the current ranges of both biological species (n = 191) and phylogroups (n = 68), and concluded that both diversification and extinction rates are higher at more northerly latitudes (for both categories). To test this conclusion, we used our data set of 260 avian sister lineages (both species and phylogroups) to construct a plot of genetic distance (a proxy for divergence time) vs. the latitude of the midpoint of the current range for comparison.