Avaliação e Modelagem da Evapotranspiração de Referência Usando Diferentes Técnicas de Aprendizado de Máquina para uma Savana Tropical Brasileira

• Ano de defesa: 2023
• Autor(a) principal: Thiago Alves Spontoni
• Orientador(a): Thiago Rangel Rodrigues
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Fundação Universidade Federal de Mato Grosso do Sul
• País: Brasil
• Palavras-chave em Português: Pantanal; mudanças climáticas; combinação; inteligência artificial; redes neurais artificiais.
• Link de acesso: https://repositorio.ufms.br/handle/123456789/6570

Meteorological elements have different influences on the environment and can significantly modify the natural vegetation development process, thus contributing to climate change. This study investigated the use of meteorological variables to determine Reference Evapotranspiration (ETo) in a transition region between the Cerrado and the Pantanal. The aim was to apply machine learning techniques to determine ETo with as few variables as possible. The results indicate that the application of artificial intelligence can promote substantial improvements in environmental modeling when alternative forecasting techniques are employed, resulting in reduced project costs and more reliable results. This research sought to identify the most efficient combination of machine learning techniques, such as Artificial Neural Networks, Random Forest and Support Vector Machines. As a result, a new model was developed, which depends on a smaller number of climatic variables compared to the Penman-Monteith method (the standard method for estimating reference evapotranspiration) and which manages to efficiently describe ETo. Machine learning techniques have proved highly effective in this type of modeling, due to their ability to process large volumes of data and identify the best interactions between the parameters involved. Furthermore, when artificial neural networks are used, an accuracy of over 94% was obtained in determining ETo when using a reduced number of variables compared to the standard method.