Use of Artificial Neural Networks in precipitation forecasting of rainy season

• Autor(a) principal: Dantas, Daniel
• Data de Publicação: 2016
• Outros Autores: Luz, Tarço Murilo Oliveira, Souza, Maria José Hatem de, Barbosa, Gabriela Paranhos, Cunha, Eduarda Gabriela Santos
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Revista Espinhaço
• Texto Completo: https://revistas.ufvjm.edu.br/revista-espinhaco/article/view/82
• topic: Cenários climáticos; modelagem do clima; previsão do tempo; Climatic sceneries; climate modeling; weather forecasting

This study aims to estimate the precipitation in the rainy season in Diamantina (MG) based on precipitation of dry previous seasons, by using Artificial Neural Networks (ANN). The chronological order of the data was changed so that the dry season of and year was related to the rainy season of the next year. A part of the data was used in the ANN training and other part used to evaluate the performance of it. The used analysis was time series and the best network found was of radial basis function type. The ANN found showed an average of error of 10%. The average precipitation of the period used in application of the network was of 1099 mm, while the average estimation was 1128 mm. The use of dry season’s data to estimate the precipitation of the rainy season presented satisfactory results and, the change of the chronological order of the dry period result in a neural network with more effective forecasting despite the unchanged one.