Antoniassi et al. (2026) Level Prediction of Rivers in the Hydrographic Region of the Paraguay River Using Machine Learning Algorithms

Identification

• Journal: Lecture notes in networks and systems
• Year: 2026
• Date: 2026-01-01
• Authors: Rogério Alves dos Santos Antoniassi, Carlos Roberto Padovani, Renato Porfírio Ishii
• DOI: 10.1007/978-3-032-10721-3_78

Research Groups

• Instituto Federal de Educação, Ciência e Tecnologia de Mato Grosso do Sul, Três Lagoas, MS, Brazil
• Empresa Brasileira de Pesquisa Agropecuária - Embrapa Pantanal, Corumbá, MS, Brazil
• Universidade Federal de Mato Grosso do Sul - Faculdade de Computação, Campo Grande, MS, Brazil

Short Summary

This study investigates the application of Machine Learning (ML) techniques for predicting river levels in the Paraguay River Hydrographic Region (RH-Paraguay). It found that Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional Long Short-Term Memory (BiLSTM) models significantly outperform the currently used Regression technique, with GRU demonstrating the highest accuracy.

Objective

• To investigate the application of Machine Learning (ML) techniques for predicting river levels in the Paraguay River Hydrographic Region (RH-Paraguay).

Study Configuration

• Spatial Scale: Paraguay River Hydrographic Region (RH-Paraguay), specifically the Pantanal, using data from three selected river stations.
• Temporal Scale: Daily river level values.

Methodology and Data

• Models used: Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Bidirectional Long Short-Term Memory (BiLSTM), and a traditional Regression technique (for comparison).
• Data sources: A dataset comprising daily river level values collected from three selected stations within the RH-Paraguay.

Main Results

• All three Machine Learning models (LSTM, GRU, BiLSTM) demonstrated improved performance compared to the currently used Regression model for river level prediction.
• The GRU algorithm model achieved the lowest error rates and was 23.84% more accurate than the Regression model.
• The LSTM model was 18.09% more accurate than the Regression model.
• The BiLSTM model was 19.16% more accurate than the Regression model.
• The LSTM and BiLSTM models showed a closer approximation to actual values during peaks of maximum and minimum river levels.

Contributions

• This study provides a quantitative demonstration of the superior performance of advanced Machine Learning models (LSTM, GRU, BiLSTM) over traditional Regression for river level prediction in the hydrologically complex Paraguay River Hydrographic Region.
• It identifies GRU as the most accurate model among those tested for overall prediction, while highlighting LSTM and BiLSTM for their ability to better approximate extreme (peak) river levels.
• The findings offer valuable insights for improving flood forecasting and water resource management in the Pantanal and similar flood-prone regions.

Funding

• Not explicitly mentioned in the provided text.

Citation

@article{Antoniassi2026Level,
  author = {Antoniassi, Rogério Alves dos Santos and Padovani, Carlos Roberto and Ishii, Renato Porfírio},
  title = {Level Prediction of Rivers in the Hydrographic Region of the Paraguay River Using Machine Learning Algorithms},
  journal = {Lecture notes in networks and systems},
  year = {2026},
  doi = {10.1007/978-3-032-10721-3_78},
  url = {https://doi.org/10.1007/978-3-032-10721-3_78}
}