Scholarlink

Water scarcity has emerged as a critical issue in Mexico, with water quality and management undergoing constant changes due to natural phenomena and human activities. Recent advancements in machine learning have demonstrated promising results in forecasting climatological and hydrological variables. This study employs a Long Short-Term Memory (LSTM) recurrent neural network to correlate 52 years of historical rainfall and streamflow data from the Upper Basin of the Guayalejo-Tamesi River in Tamaulipas, Mexico, aiming to provide accurate flow forecasts through the rainfall historical data. The results show that LSTM produced a Root Mean Squared Error (RMSE) of 110.44. We acknowledge various factors like data cleaning processes such as imputation and outlier removal, the accuracy of historical data, and unconsidered climatological variables as potential influences on the network’s performance.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

LSTM Recurrent Neural Network for Streamflow Forecasting in the Upper Basin of the Guayalejo-Tamesí River, Mexico

  • 章节

  • 发表日期：

    2026年

查看全文
  • Estefanía Román-Rangel,
  • Jesús David Terán-Villanueva,
  • Tomás Alejandro Peña-Alonso,
  • Salvador Ibarra-Martínez,
  • Mirna Patricia Ponce-Flores

摘要

Water scarcity has emerged as a critical issue in Mexico, with water quality and management undergoing constant changes due to natural phenomena and human activities. Recent advancements in machine learning have demonstrated promising results in forecasting climatological and hydrological variables. This study employs a Long Short-Term Memory (LSTM) recurrent neural network to correlate 52 years of historical rainfall and streamflow data from the Upper Basin of the Guayalejo-Tamesi River in Tamaulipas, Mexico, aiming to provide accurate flow forecasts through the rainfall historical data. The results show that LSTM produced a Root Mean Squared Error (RMSE) of 110.44. We acknowledge various factors like data cleaning processes such as imputation and outlier removal, the accuracy of historical data, and unconsidered climatological variables as potential influences on the network’s performance.