Influência da extensão espacial sobre modelos de adequabilidade de habitat para primatas da Mata Atlântica

Abstract

Habitat suitability modeling is a tool that analyzes the relationships between the species occurrence and environmental variables. However, the definition of the spatial extent used in the construction of models can influence the results of the models. Primates are the group of mammals most threatened with extinction in the world, and the Atlantic Forest is an important biome for the conservation of this group. In this study, we evaluated how the definition of spatial extension influences the suitability of habitat suitability models for 14 primate species of Atlantic Forest. We built models in MaxEnt using three extents: i) general extension, covering the occurrence records of all species studied; ii) study area extension (Atlantic Forest); iii) specific extension (occurrence extent of the species). The suitability pattern remained the same in the three approaches for all species, but the intensity of suitability varied. Models constructed in larger extensions had high frequency in extreme values and models constructed using smaller extensions had higher frequency in intermediate values. Models constructed on smaller extensions capture less environmental variability, which probably explain this result. Pixel suitability values differed significantly among extents for all species. The values of the differences among the suitability of the models were predominantly positive and close to zero, however, the spatial pattern of the surfaces of differences of suitability between the models remained the same. Bigger differences predominated between models generated in general and specific extensions. Models constructed in larger extensions are more susceptible to commission errors, however they allow better prediction of absence areas. AUC values were higher in the general and Atlantic Forest extensions. Models constructed on large extensions result in a higher absence rate and overestimate the prediction of presence, increasing AUCs and commissions’ errors. Although the spatial pattern has not changed, the difference in suitability frequency can have consequences when the suitability surfaces are converted into geographic ranges taking into account a limit of suitability value (threshold).