Estimación de caudales en una cuenca tropical mediante modelación neuronal artificial basada en memoria hidrológica.

Saved in:

Bibliographic Details
Title:	Estimación de caudales en una cuenca tropical mediante modelación neuronal artificial basada en memoria hidrológica.
Alternate Title:	Estimating flow rates in a tropical basin using artificial neural modeling based on hydrological memory.
Authors:	Darío Blandón-López, León¹ lblandon@unal.edu.co, González-Salcedo, Luis Octavio²
Source:	Ingeniería y Competitividad. ene-mar2026, Vol. 28 Issue 1, p1-16. 32p.
Subjects:	Long short-term memory, Hydrological forecasting, Hydrologic models, Southern oscillation, Artificial neural networks, Water management, Climate change, Watersheds
Geographic Terms:	Valle del Cauca (Colombia)
Abstract (English):	Introduction: Hydrological systems have been severely impacted by the increase in extreme events such as floods and droughts, intensified by climate change. The Tuluá River basin, located in the Cauca Valley of Colombia, faces additional challenges due to deforestation and limited monitoring coverage, complicating efficient water resource management. Objectives: Develop a neural model based on LSTM networks to predict the flow rate of the Tuluá River from hydrometeorological variables, for the anticipation of extreme scenarios and support in decision-making in regional water management. Methodology: 6875 daily records for five variables (precipitation, temperature, evaporation, relative humidity, and streamflow) from seven stations were consolidated. To fill missing values, data preprocessing was applied using KNN and Pearson correlation analysis. The LSTM model was configured and trained, and evaluated for the training, test, and validation subsets using standard neural network metrics. Results: The neural model demonstrated high predictive capacity with correlation coefficients R (0.75-0.98). It was evident that the model effectively captures temporal dependencies and nonlinear patterns, in addition to identifying the influence of the ENSO phenomenon on local hydrological dynamics. Conclusions: The LSTM model is an effective tool for estimating flows for the basin under study, facilitating both water management and the prevention of risks associated with climate variability. Future developments can enhance its accuracy and usefulness if they include global climate indices. [ABSTRACT FROM AUTHOR]
Abstract (Spanish):	Introducción: Los sistemas hidrológicos han sido severamente impactados por el aumento de eventos extremos como inundaciones y sequias, a su vez intensificados por el cambio climático. La cuenca del río Tuluá, ubicada en el Valle del Cauca - Colombia, enfrenta retos adicionales por la deforestación y limitada cobertura de monitoreo, complicando la gestión eficiente del recurso hídrico. Objetivos: Desarrollar un modelo neuronal basado en redes LSTM para predecir el caudal del río Tuluá a partir de variables hidrometeorológicas, para la anticipación de escenarios extremos y el apoyo en la toma de decisiones en la gestión hídrica regional. Metodología: 6875 registros diarios para cinco variables (precipitación, temperatura, evaporación, humedad relativa y caudal) de siete estaciones, fueron consolidadas. Para completar valores faltantes se aplicó preprocesamiento de datos usando KNN y análisis de correlación de Pearson. Se configuró y se entrenó el modelo LSTM, el cual se evaluó para los subconjuntos de entrenamiento, prueba y validación mediante métricas usuales para redes neuronales. Resultados: Una alta capacidad predictiva con coeficientes de correlación R (0,75 - 0,98) mostró el modelo neuronal. Se evidenció que el modelo captura efectivamente las dependencias temporales y patrones no lineales, además de identificar la influencia del fenómeno ENSO en la dinámica hidrológica local. Conclusiones: El modelo LSTM es una herramienta efectiva en la estimación de caudales para la cuenca en estudio, lo cual facilita tanto la gestión hídrica como la prevención de los riesgos que se asocian a la variabilidad climático. Futuros desarrollos pueden potenciar su precisión y utilidad, si consideran la inclusión de índices climático globales. [ABSTRACT FROM AUTHOR]
	Copyright of Ingeniería y Competitividad is the property of Universidad del Valle and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	Introduction: Hydrological systems have been severely impacted by the increase in extreme events such as floods and droughts, intensified by climate change. The Tuluá River basin, located in the Cauca Valley of Colombia, faces additional challenges due to deforestation and limited monitoring coverage, complicating efficient water resource management. Objectives: Develop a neural model based on LSTM networks to predict the flow rate of the Tuluá River from hydrometeorological variables, for the anticipation of extreme scenarios and support in decision-making in regional water management. Methodology: 6875 daily records for five variables (precipitation, temperature, evaporation, relative humidity, and streamflow) from seven stations were consolidated. To fill missing values, data preprocessing was applied using KNN and Pearson correlation analysis. The LSTM model was configured and trained, and evaluated for the training, test, and validation subsets using standard neural network metrics. Results: The neural model demonstrated high predictive capacity with correlation coefficients R (0.75-0.98). It was evident that the model effectively captures temporal dependencies and nonlinear patterns, in addition to identifying the influence of the ENSO phenomenon on local hydrological dynamics. Conclusions: The LSTM model is an effective tool for estimating flows for the basin under study, facilitating both water management and the prevention of risks associated with climate variability. Future developments can enhance its accuracy and usefulness if they include global climate indices. [ABSTRACT FROM AUTHOR]
ISSN:	01233033
DOI:	10.25100/iyc.v28i1.15199