Ciência de dados aplicada à conservação e à bioprospecção marinha: síntese do banco de dados do SISBIO para Arraial do Cabo (RJ) e Fernando de Noronha (PE), Brasil

Abstract

The Marine Extractive Reserve of Arraial do Cabo/RJ (ResexMar-AC) and the Marine National Park of Fernando de Noronha/PE (ParnaMar-FN) are the two Federal Marine Conservation Units with the largest tourist flow in Brazil For the past 10 years, the results of the research conducted in these areas has been stored by the SISBio online system. Here we propose a synthesis methodology for the vast amount of reported information received and accumulated in this database. We developed a tool (BioMining) to improve the search and data processing, which is also used as support to machine learning algorithms with supervised text classification models in the context of SISBio textual data, which were tested and investigated. The reported information were classified into seven categories: 'Taxa', 'Location', 'Impacts', ‘Inspection’, 'Monitoring', 'Future researches', and 'Management actions'. The linear model SVM (Support Vector Machine) and the ensemble RFC (Random Forest Classifier) obtained the best accuracy in the classification of the texts, reaching the maximum of 84%, and the categories with less precision were 'Inspection' and 'Impacts', due to the smaller amount of samples. ParnaMar-FN presented a significantly larger number of reports and mean number of words, and this stimulated an increase in the accuracy of the text classification models. In addition, data on the biodiversity of the protected areas suggest that ResexMar-AC has a richer marine environment (corroborating with the literature), with emphasis on sessile invertebrates, while in ParnaMar-FN the surveys focus on the megafauna (dolphins, fishes and turtles). Available literature suggests that the species listed in the ResexMar-AC have a greater biotechnological potential, highlighting exotic species and native species considered exotic in other parts of the world. The bioinvasion presented itself as quite relevant in both units, being considered as impacting and a necessary target for monitoring and future researches, principally the marine invertebrates at Anjos Bay in Arraial do Cabo, and the terrestrial vertebrates in Fernando de Noronha Archipelago. Tourism also presented itself as one of the activities most reported as impacting and requiring inspection and planning in the units, after the fisheries activities in ResexMar-AC, as expected, due to the diversity of extractive activities in this unit. The results of this thesis enhances the potential in filling the gaps between scientific information and decision making, highlighting some knowledge gaps and the potential of data science in the field of marine biotechnology and conservation.