Modelos de distribuição de espécies de vellozia (Velloziaceae) endêmicas da Cadeia do Espinhaço e o efeito amostral sobre os mapas preditivos

Abstract

The effectiveness of conservation actions depends on the knowledge of the geographical distribution of species. However, this knowledge is far from being achieved for most species, especially those occurring in mountainous tropical environments, such as the rocky fields (campos rupestres) of the Espinhaço Range, characterized by forming islands of vegetation in the higher parts of mountains isolated by altitudinal variation. This discontinuity and difficulty of access aggravate the lack of knowledge about the distribution of species, which would be essential for studying the evolutionary and ecological processes in these environments. Vellozia auriculata and V. gigantea are examples of this problem as they were initially known by the occurrence in only one or a couple of hills. New records have been found in survey efforts targeted for these species. Species distribution models (SDMs) have proved to be a useful tool to predict the distribution of species and guide field research in order to find new records. The aim of this study was to use the algorithm Maxent to locate new populations of endemic species in the Espinhaço Range and to assist actions to expand, creation and management of protected areas, in addition to quantify the environment range of the occurrences of these species. The SDMs associate species occurrence data and environmental information available to set appropriate conditions where populations can be maintained, and extrapolate the distribution of species in geographic space. The quantity and quality of occurrence data are the crucial points to a successful outcome prediction models and application to conservation. However, beyond scarce, most data are biased resulting in maps with high degree of uncertainty. Therefore, we tested the influence of sample size and the spatial bias in the prediction of different algorithms (Bioclim, Domain, Environmental Euclidean Distance and Maxent) in these environments marked by great topographical, geological and environmental diversity, using these two species as an example. New locations were found for both species, seven for V. auriculata and five for V. gigantea. Due to anthropogenic threats, potentiality of presenting a genetical divergence, especially in extremes of distribution, and the classification of "Endangered", these species need urgent conservation and management. In this study, the maps generated by all algorithms, except for BIOCLIM, from small sample x numbers generated satisfactory prediction for use in exploratory actions to guide field surveys, with larger samples generating more accurate maps. Domain was generally less sensitive to the influence of sample size and less influenced by the bias space, showing one of the highest AUC values, and almost identical to the maps generated from optimal sampling, including all the occurrence data. New sampling conducted in the field from the known records of the species was essential to increase the accuracy of the maps and find new records in the field. Because this is the first study using this methodological approach in the rocky fields, the results obtained with both the application and the evaluation of methods may serve as a guide for future studies of species distribution in the Espinhaço Range, and may also be used to comparison between different approaches of applicability and methodological evaluation in species distribution modeling.