Spatial distribution models
Excluding the case of the striped owl (Asio clamator ), the empirical entrograms showed no substantial variations when excluding those localities closer than 25 km (Supplementary Material A, Appendix 1. Fig. C). Thus, we performed the SDMs including all the occurrences, except for the aforementioned species.
We provide detailed information of the models in Supplementary material A, Appendix 3, Table A. In concordance with Liu et al. (2005, 2016), we found that thresholds based on sensitivity-specificity outperformed the remnant ones (Supplementary material Appendix 1, Fig. D). For such a reason, we used the sensitivity-specificity equality threshold to create the binary maps.
The climatic variables were usually the most important predictors of suitability (Table 3). Temperature-based PC1 and, especially, PC4 ranked highest for percentage contribution for 27 of the taxa studied, followed by precipitation-based PC3 (nine taxa), geology (four), and both soil and PC2 (one each).
The predictions of the monotypic species fitted the best their traditionally reported distributions and had lower omission percentages (median 13%, ranging between 0% for G. mooreorum , to 31%, forLophostrix cristata ; Supplementary material A, Appendix 1, Fig. E) compared to the polytypic ones (median 34%, from 24% forStrix virgata to 42% for Megascops choliba ). These omissions usually felt outside the corresponding most represented biome (often, outside the Atlantic Forest). Besides, we found a general tendency towards fitting improvements after running models based on occurrences of their respective subspecies (median 15% of omissions, from 4% in Athene cunicularia cunicularia and, exceptionally, 66% for Strix huhula huhula ). For the endemic and probably extinct Pernambuco pygmy-owl, our models predicted a very restricted range around both known localities, but also two additional separated spots, one located in the protected area of Manguezais da Foz do Rio Mamanguape, and the other in the mouth of the Sergipe river: unassessed areas from the ornithological point of view.
According to the sensu stricto map (Fig. 2A), the Atlantic Forest hosts the highest potential richness (ca. 15 species), especially within the Dense Ombrophylous Forest range (around the littoral and mountainous areas of the Southeastern region). Scattered areas along the Amazonas river lowlands, notably in the belt of siliciclastic sedimentary rocks north of the river and around its mouth are also highly diverse. Conversely, wide coldspots (around zero predicted species) characterize more open environments such as Cerrado, Caatinga or Pampa, as well as broad areas in the Amazonia. Moreover, the sensu lato map (Fig. 2B) keeps the same areas of high biodiversity (over 15 taxa) but reducing the extension of the coldspots in the Amazonia, keeping only some areas in the Rondônia State. Thus, both approaches indicate that the Atlantic Forest, which harbors the highest richness, is poorly covered by strictly protected areas since these become substantially smaller and sparser within a gradient from Northwest to Southeast Brazil. However, by comparing the number of species (sensu stricto ) recorded against those predicted, we found that all biomes are under-sampled (Table 2), especially the Pantanal and the Caatinga.