We here describe two new species endemic to the state of Espírito Santo, Brazil, that belong to different clades within Miconia s.lat. Miconia quartzicola is presumably part of the Leandra s.str. clade, as indicated by its terminal inflorescences and petals with an acute apex. It has been collected only once in the municipality of Vargem Alta in disturbed vegetation on loose quartzitic substrate (‘morros de sal’). Miconia spiritusanctensis belongs to Miconia sect. Cremanium, as indicated by its small white and obovate anthers with four apical pores. The latter encompasses populations previously identified as M. hirtella that are morphologically, geographically and climatically segregated from typical populations of the species from more dry and inland regions of Brazil. In addition to the descriptions of the new species, we present comments, conservation status and plates for both, as well as climatic modelling analyses on the populations of M. spiritusanctensis and M. hirtella. We recommend that Miconia quartzicola and M. spiritusanctensis should both be considered as threatened, ‘critically endangered’ and ‘endangered’, respectively, according to the IUCN extinction risk criteria.

In order to compare climatic preferences of M. hirtella and M. spiritusanctensis, we gathered data on their distribution from online data available in the biodiversity portal SpeciesLink (<http://splink.cria.org.br/>). The resulted database was then filtered in several ways. Briefly, we considered only identifications by specialists and, whenpossible (image available), we checked all the other identifications. All specimens lacking images or reliable identifications were removed. We also deleted all material without detailed location, but for the ones with a detailed description of the locality, despite lacking coordinates, we extracted the coordinates using Google Earth based on the label information. In order to reduce putative occurrence bias on the models, the data set was spatially thinned with the R package spThin ver. 0.1.0 (Aiello-Lammens et al. 2014). Only points with a minimum distance of 5 km apart from each other were used for further analyses. The potential distribution of the taxa under current climatic conditions were modeled and evaluated by Maxent ver. 3.4.0 (Phillips and Dudík 2008) within the R package dismo (Hijmans et al. 2017). Climatic models were based on the 19 bioclimatic layers of the WordClim data set (Hijmans et al. 2005) under current conditions (30″ special resolution). For each taxon, a mask was created with a buffer of 1000 km of diameter around its known distribution. Additionally, to exclude the immediate area around the known localities from the background, a buffer of 100 km in diameter was generated for each known point and subtracted from the main mask. We extracted the values for all bioclim layers of each coordinate and calculated the mean of each species (Reginato and Michelangeli 2019). The area under the curve (AUC) of the receiver operating characteristic (ROC) was used as evaluation criterion. Climatic tolerances were compared between the species through climatic envelope profiles. The records were intersected to the layers using the R package raster (Hijmans 2020). The extracted climatic values were summarized with a principal component analysis (PCA) using the R package ade4 ver. 1.7.6 (Dray and Dufour 2007), and convex hulls were plotted per group under comparison (Reginato and Michelangeli 2019). Values of elevation of each specimen were extracted through the R package elevatr ver. 0.3.1 (Hollister 2020) and the median was calculated for both taxa.