Modelagem de distribuição potencial de espécies como ferramenta para a conservação: seleção e avaliação de algoritmos e aplicação com Heliconius nattereri Felder, 1865 (Nymphalidae: Heliconiinae)

• Autor(a) principal: Barreto, Francisco Cândido Cardoso
• Data de Publicação: 2008
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: LOCUS Repositório Institucional da UFV
• Texto Completo: http://locus.ufv.br/handle/123456789/867
• topic: Modelagem de distribuição potencial de espécies; MAXENT; GARP; SVM; Avaliação de modelos; Heliconius nattereri; Tetrastylis ovalis; Species conservation; Potential distribution modelling; MAXENT; GARP; SVM; Models evaluation; Heliconius nattereri; Tetrastylis ovalis; CNPQ::CIENCIAS BIOLOGICAS::ECOLOGIA::ECOLOGIA TEORICA

This thesis is divided into two chapters. The first one presents a technical approach in which with an artificial created species, I ve tested three different algorithms used to predict the species potential geographic distribution: MAXENT, GARP and SVM. The first chapter s objective was to test the performance of these three algorithms and determine which one is the best to predict the real distribution of the artificial species. Because the performance of the models can be evaluated by many techniques, I ve tested which metric presents the most real results based on the errors and successes rates, calculated directly by the comparison of each model prediction in relation to the real occurrence area of the artificial species. Besides, use these metrics, it is necessary the definition of a threshold that divides the model in two aspects: 1- Above this threshold, the omission and commission errors are balanced; and 2- Below this threshold the models have an error rate bigger than the acceptable. Knowing the importance of this definition, I used two different approaches find in the literature: the threshold defined with the construction of a ROC curve and the minimum value of environmental suitability attributed, by the model, to a real presence point. The results indicate that this definition is very important to evaluate correctly each model. When the ROC threshold was used, the algorithm with the best performance was GARP. When the Minimum threshold was used, this situation changed and MAXENT presented best performance. The second chapter uses the knowledge generated by the first chapter and adopts MAXENT to model the potential geographic distribution of the butterfly Heliconius nattereri Felder, 1865, classified as vulnerable in the IBAMA endangered species list, published in 2003. This species has occurrence registries in Minas Gerais, Espírito Santo e Bahia, and presents the characteristics of being monophagic about the resource of its larvae the species of the Passifloraceae family known as Tetrastylis ovalis (Vell.) Killip. This passion- fruit species has low vegetative growing and is pollinated by bats, occurring in a relative restrict distribution in Brazil, and with local low abundance. Because of this strict dependence relation of the butterfly with its larval resource, the potential geographic distribution of Tetrastylis ovalis was used as a reference to the refining of the potential distribution of H. nattereri. With the results, the prediction of the environmental suitability degree to the species occurrence in Brazil, I was able to cross this data with the conservation units information, making a list of the conservation units that have a chance of being occupied by unknowing populations of H. nattereri. I also discuss some actions that can be made in order to reduce the risk level of this species, by increasing its registry local populations and finding new ones in different places, such as: (1) Growth and management of H. nattereri s host plant, T. ovalis, increasing its abundance and availability in the protection areas; (2) Population dynamics studies of the species; (3) Genetic variability studies of the species; (4) Individuals translocation, if the necessity is determined in item (3); (5) H. nattereri captive creation in Butterfly Houses in the own conservation unit, with the basic information obtained in (2) and (3) items.