The Prediction of Hydrometeorology Variables Using the Method of Recurrent Neuronal Networks

• Autores: Díaz León, J. A., Olarte Escobar, M. A., Jara García, M.
• Formato: artículo
• Fecha de Publicación: 2023
• Institución: Universidad Peruana de Ciencias Aplicadas
• Repositorio: UPC-Institucional
• Lenguaje: inglés
• OAI Identifier: oai:repositorioacademico.upc.edu.pe:10757/668740
• Enlace del recurso: http://hdl.handle.net/10757/668740
• Nivel de acceso: acceso embargado
• Materia: Flow prediction; Lima metropolitana; Long-term bidirectional recurrent neural network (BRNN); Precipitation prediction; Temperature prediction; Hydrometeorological variables; Temperature, precipitation, and flow; Bidirectional recurrent neural networks (BRNN); Metropolitan Lima; LSTM models; Mean square error (MSE); Root mean square error (RMSE); Nash Sutcliffe efficiency ratio (NSE); Average temperatures; Prediction performance

This scientific article evaluates the prediction of hydrometeorological variables, which refer to temperature, precipitation, and flow. The applied methodology is long-term bidirectional recurrent neural networks (BRNN) in a series of 40 years of study for a better perspective on the climatological conditions in Metropolitan Lima until the year 2050. The BRNN model is formed by a single series of past observations, which means that the model analyzes one variable simultaneously to project the next value in the sequence, and unlike other LSTM models, the bidirectional model can model complex and long-time series of sequences efficiently. The purpose of the model is to divide the neurons of a regular RNN into 2 directions, one of them is for the positive time direction (forward states), and one is for the negative time direction (reverse states). In addition, mean square error (MSE), root mean square error (RMSE), and Nash Sutcliffe efficiency ratio (NSE) were used as error metrics to assess prediction performance. Regarding the results, the prediction of average temperatures tends to increase between the ranges of 0.30–0.90 °C with estimated maximum temperatures up to 27 °C.