Aprendizado de máquina aplicado à teoria do nicho para soluções em ecologia e conservação: um tratado sobre Lonchophylla bokermanni Sazima, Vizotto & Taddei, 1978 (Chiroptera,Phyllostomidae)

Abstract

Ecological niche modeling and species distribution modeling are conceptually distinct tools, grounded in Niche Theory, that aim to implement machine learning algorithms to establish equivalences between niches and habitats. This has facilitated the development of a new science, with the autonomy to test hypotheses in the fields of biogeography and conservation biology. The present thesis focuses on the application of informatics to theoretical ecology to provide solutions for some of the challenges involved in species conservation, using the nectarivorous bat Lonchophylla bokermanni as a study model. This species is rare, "endangered" in nature, and has a strict relationship with endemic bromeliads in the Serra do Espinhaço Biosphere Reserve (Encholirium subsecundum and E. vogelii). The first chapter outlines the identification of a previously unknown species within the genus Lonchophylla through ecological niche modeling. The second chapter provides an initial synthesis of the challenges in incorporating biotic factors into species distribution models, until an approach was developed and tested. The third and final chapter discusses the prospecting of new occurrences for L. bokermanni directed by species distribution modeling. The data collected throughout this thesis are relevant not only for refining the design of biodiversity conservation strategies within the Serra do Espinhaço Biosphere Reserve, but also for providing a theoretical framework for future studies on the ecology of L. bokermanni, E. subsecundum, and E. vogelii.