Hydrological modeling using artificial neural networks for flood event forecasting. Case study: Pomba river in Santo Antônio de Pádua - RJ

Flood prediction through hydrological modeling of watersheds remains an emerging need in society, particularly in regions highly affected by these extreme events. Models based on artificial neural networks have demonstrated significant potential for addressing this issue due to their simplicity and...

Descripción completa

Detalles Bibliográficos
Autores: Dias, Rennan Mendes de Moraes dos Santos, Telles, Wagner Rambaldi, Silva Neto, Antônio José da
Tipo de recurso: artículo
Estado:Versión publicada
Fecha de publicación:2024
País:Brasil
Institución:Universidade Federal de Santa Maria (UFSM)
Repositorio:Revista Ciência e Natura (Online)
Idioma:inglés
OAI Identifier:oai:ojs.pkp.sfu.ca:article/87221
Acceso en línea:https://periodicos.ufsm.br/cienciaenatura/article/view/87221
Access Level:acceso abierto
Palabra clave:Artificial neural networks
Hydrological modeling
Flood event
Multilayer perceptron
Redes Neurais Artificiais
Modelagem Hidrológica
Eventos de Cheias
Perceptron Multicamadas
Descripción
Sumario:Flood prediction through hydrological modeling of watersheds remains an emerging need in society, particularly in regions highly affected by these extreme events. Models based on artificial neural networks have demonstrated significant potential for addressing this issue due to their simplicity and agility. In this study, a model was developed using a multilayer perceptron network for predicting river discharge and water level based on the previous day's river state and precipitation forecast. The Pomba river in the city of Santo Antônio de Pádua-RJ was investigated due to its regular occurrence of flood events that impact the entire population. Metric and graphical results showed the model's strong ability to estimate discharge and water levels throughout the year at a station with limited data. On the other hand, the model encountered difficulties in accurately estimating peak values.