Banca de DEFESA: DANILO SILVA LISBOA
Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.DISCENTE : DANILO SILVA LISBOA
DATA : 14/09/2020
HORA: 14:00
LOCAL: Virtual: Google Meet
TÍTULO:
Probabilistic forecast of coral bleaching: theoretical and proactical aspects for the development of an early alert system to reefs of the Atlantic Ocean
PALAVRAS-CHAVES:
Coral bleaching forecast; Bayesian Network Model; Short-term prediction; Atlantic Ocean; Early warning system
PÁGINAS: 110
GRANDE ÁREA: Ciências Exatas e da Terra
ÁREA: Oceanografia
SUBÁREA: Oceanografia Geológica
ESPECIALIDADE: Geofísica Marinha
RESUMO:
Coral bleaching represents the most prominent negative response of reef ecosystems to current climate change. Capable of promoting mass mortality of colonies and significant changes in reef community structures, these phenomena should become increasingly frequent, intense, long-lasting and globally widespread, to the point of compromising the viability of reef ecosystems in the coming decades. The choice of Bayesian nets for bleaching modelling comes from their inherent ability to integrate data and knowledge with different characteristics and to deal with the uncertainties and complexities related to these phenomena through the use of probabilities. In this research reasons why the Bayesian approach can be quite proficient for modelling reef environments and particularly for the case of coral bleaching are presented, and the most relevant aspects of key references are discussed in more detail from the perspective of the reef scientist/modeler. We describe the nature and procedures for building these models, providing recommendations for appropriate modeling in each of the major steps. We present a case study of formalizing the conceptual model of seasonal coral bleaching prediction from a control case in an environmental reserve area. For this, two well founded scientific ideas have been evaluated and confirmed in relation to their forcefulness in bleaching events: 1- there is a relationship between the intensity of the El Niño phenomenon and positive thermal anomalies in the Atlantic Ocean; and 2- the bleaching of corals is mainly influenced by positive thermal anomalies in seawater. The formalised conceptual model has hyeraquically organised the indicators related to these assumptions into a network structure according to their levels of influence on bleaching and has been used as a starting point to develop competing models capable of making seasonal forecasts for the reef areas of the North Atlantic Ocean. Data mining procedures, validation, and scoring tests were used as criteria for comparison, demonstrating the feasibility of the Bayesian approach to make seasonal forecasts of the bleaching state with accuracy levels above 80%. As alternatives to improve accuracy, customized models were developed with databases restricted to specific situations, but at the cost of loss of predictive capacity. We believe that the Bayesian network model developed, tested and evaluated in this research represents a useful resource to assist scientists, governments and environmental managers through a bleaching early warning system capable of providing sufficient time to plan field campaigns and possible mitigation actions. Finally, from the point of view shown in this research, the Bayesian approach represents an alternative with expressive potential to assist the management of reef ecosystems and should be established as a standard technique of analysis in the coming years. Familiarizing oneself with the characteristics and procedures of this approach should be of great value to students/researchers and managers committed to safeguarding the future viability of these priceless ecosystems.
MEMBROS DA BANCA:
Presidente - 1348177 - RUY KENJI PAPA DE KIKUCHI
Interno(a) - 632.336.575-87 - ZELINDA MARGARIDA DE ANDRADE NERY LEAO - UFBA
Externo(a) ao Programa - 2130353 - RICARDO ARAUJO RIOS
Externo(a) à Instituição - ESTEVAM RAFAEL HRUSCHKA JUNIOR
Externo(a) à Instituição - MARILIA DE DIRCEU MACHADO DE OLIVEIRA