Sampling biases, distribution patterns and taxonomic knowledge
Brazil harbors a rich owl diversity, distributed unequally throughout its territory. Indeed, our models suggest a markedly variable richness across the Country, ranging between zero and 16, averaging three, species per 30 arcsec * 30 arcsec pixel. Since 69% of the occurrences are from Atlantic Forest and Amazonia, their higher predicted richness can partially respond to sampling biases. The Atlantic Forest exemplifies those situations where sampling efforts are favored by neighboring established research centers and major urban areas (Brito et al., 2009; Moerman & Estabrook, 2006; G. V. T. Ribeiro et al., 2016). The Amazonia replicates the pattern already described for botanical and ornithological records (Vale & Jenkins, 2012), probably prompted by local facilities or recurrent bird-oriented inventories.
However, all biomes are under-sampled, as suggested by the difference between the number of recorded and predicted species, especially in the Caatinga and Pantanal. Hence, our results agree with Silva (1995) and Fernández-Arellano et al. (2021) regarding the need for research efforts addressed to unexplored areas and periods, less detectable species, in neglected biomes. It is reasonable to expect that such efforts oriented to nocturnal birds would increase the respective species lists for all biomes, especially in Caatinga and Pantanal.
Excluding the possibility of factors shaping the distributions of the Brazilian Strigidae, not considered in the present study, several not mutually exclusive possibilities can explain the overall elevated number of omissions for the models. First, misidentifications. For example, (Rocha & López-Baucells, 2014b) reported a young Lophostrix cristata that was indeed a Strix virgata (Rocha & López-Baucells, 2014a). Thus, similar unnoticed mistakes, especially those involving species more alike, can result in false occurrences reported far from the actual range.
Second, vagrancy, a phenomenon frequently reported in the ornithological literature at least since Grinnell (1922), including cases of owls impacting native fauna in remote islands (Bried, 2003), and prompted by factors as diverse as expanding populations, weather conditions, geography, age, and genetics (Kalwij et al., 2019; Ralph & Wolfe, 2018; Veit, 2000). According to the ‘Exodus Hypothesis’ (Flade & Lachmann, 2008), vagrancy can be triggered by severe habitat loss, and individuals occasionally can settle in available but less adequate habitats affecting their fitness and population permanence (Pärt et al., 2007; Robertson & Hutto, 2006). Here, most omissions corresponded to taxa from the Atlantic Forest recorded in open areas such as the Cerrado. Currently, the Atlantic Forest covers less than 16% of the original extent, and more than 80% of the fragments are smaller than 50 ha (Ribeiro et al. 2009; but see Rezende et al. 2018), a dramatic situation that could have triggered the translocation of owls toward the comparatively less impacted Cerrado areas in the past.
Third, range expansions, as frequently reported for Strigiformes in well-monitored areas in North America (Livezey, 2009a), Europe (Bashta, 2009; Mysterud, 2016), and Oceania (Hyde et al., 2009), occasionally mediated by human activities. For example, Livezey (2009b) postulated that the forest expansions after the arrival of European settlers (who excluded fires set by Native Americans, planted trees, and overhunted bison, elks, deers and beavers) allowed barred owls (Strix varia ) to overcome the historical lack of trees in the Great Plains of North America granting their expansion westward. According Haywood (2010), the plantations of Tasmanian Blue Gum (Eucalyptus globulus ) facilitated the range expansion of the powerful owl (Ninox strenua ) in South Australia. Similar situations could be shaping the distributions of some Brazilian owl species, such as the tree-nesterA. harrissi , whose omissions mostly correspond to records from the last two decades in open biomes such as Cerrado and Caatinga (A. F. T. da Silva et al., 2021) or Pampa (Müller Rebelato et al., 2011). That is, these new records correspond to biomes anthropized the most during the recent decades according to MAPBIOMAS (https://plataforma.brasil.mapbiomas.org/, consulted April 14, 2021), with steady increases in planted tree surfaces (http://atlasagropecuario.imaflora.org/, consulted April 16, 2021), perhaps offering new more suitable areas for these owls.
Notwithstanding, a more likely option is that the taxonomy of the Brazilian owls is far from being completely understood, with local populations/morphs representing in many cases valid species, as recently demonstrated by Dantas et al. (2021). In these cases, it is reasonable to expect SDMs biased toward the best-represented subspecies, failing to predict regions suitable for the remnant ones. Our models for the conspecific subspecies and our results on niche equivalency, even if assigning the occurrences in an approximative way, reinforced this possibility.
Our analyses agree with Peterson et al. (1999) that the speciation process involves geographic dimensions first and then ecological aspects. Environmental parameter variations may result in niche divergences (Pyron et al., 2015; Ramoni-Perazzi et al., 2020), which may explain to some extent most of the speciation patterns within the Brazilian Strigidae since our results imply that the niches of the subspecies under comparison can be more dissimilar than expected, in most cases obligated by dissimilarities of their respective backgrounds. This situation may also involve species considered as monotypic. For example, the high omission rates of over 26% in the case of the Amazonian pygmy owl (Glaucidium hardyi ) may suggest a species complex since the central portion of the Amazon River basin is a secondary contact zone for taxa isolated in the main Amazonian sub-basins during the Mid- and Late Pleistocene (Thom et al., 2020).
All the aforementioned options deserve further evaluations, especially involving morphological, molecular and bioacoustics approaches.