Hydrological forecast in Macaé river basin with neural networks

Authors

  • Julia Godinho Instituto Politécnico - UFRJ
  • Janaina Santanna Gomide Gomes Instituto Politécnico - UFRJ
  • Rafael Malheiro Instituto Politécnico - UFRJ
  • Laura Emmanuella Santana Escola Agrícola de Jundiaí - UFRN https://orcid.org/0000-0003-2086-3471

DOI:

https://doi.org/10.5335/rbca.v14i1.12964

Keywords:

Redes Neurais Artificiais, Bacia do Rio Macaé, Previsão hidrológica

Abstract

Hydrological forecasting is a valuable tool for dealing with socio-environmental problems, and it can be used in natural disaster alert systems and as assistant aid in making public policies.
This work presents an application of a hydrological model based on Artificial Neural Networks (ANN). The variable modeled was the flood stage of the fluviometric station Fazenda Airis, located in the Macaé River drainage basin.
To this end, the datasets used are composed of daily records of flow and rainfall stations between 2010 to 2013, made available by the National Water Agency (ANA) and the INEA (Environment State Institute of Rio de Janeiro) Flood Alert System.
The adopted methodology investigates the influence of the input variables and ANN architecture on the models' performance.
The results obtained were considered very satisfactory and support the proposition of the potential of Artificial Neural Networks for hydrological forecasting. It was found that of the 189 models created, 42.3 \% had the coefficient of determination R2 above 0.80.
Conclusions, The best ANN developed received daily data from six rainfall stations and one fluviometric station, obtaining for metrics R2 and MAE the values of 0.88, 7.03 cm, respectively.
Finally, the results were compared with related works and are similar or superior even with shorter time series.

Downloads

Download data is not yet available.

Downloads

Published

2022-04-17

Issue

Section

Original Paper

How to Cite

[1]
2022. Hydrological forecast in Macaé river basin with neural networks. Brazilian Journal of Applied Computing. 14, 1 (Apr. 2022), 70–80. DOI:https://doi.org/10.5335/rbca.v14i1.12964.