Scholarlink

<p>Evapotranspiration is a crucial process in hydrology, agriculture, and climate studies, significantly impacting crop yield, water resources, and climate modeling, and necessitating accurate forecasting for effective management. This research evaluates the effectiveness of deep learning and machine learning techniques in forecasting evapotranspiration (ET) by utilizing climatic factors, including Rainfall, Temperature, and Sunshine hours. The Long Short-Term Memory (LSTM) model exhibited exceptional capabilities in recognizing temporal relationships and seasonal patterns within the data, achieving a test R² of 0.66, an RMSE of 0.11, and an MAE of 0.07. Statistical significance testing showed that LSTM significantly outperformed RNN (<i>p</i> = 0.0091), and RNN outperformed GRU (<i>p</i> = 0.0324), while the performance difference between LSTM and GRU was not statistically significant (<i>p</i> = 0.2694). The bootstrap analysis revealed low standard errors for LSTM (SE = 0.0118), RNN (SE = 0.0113), and GRU (SE = 0.0114), with 95% confidence intervals indicating reliable predictions. In contrast, traditional machine learning models like Decision Tree and GBM exhibited overfitting, with high training performance but poor generalization to unseen data. The study demonstrates the effectiveness of LSTM and RNN in predicting ET, highlighting their potential in irrigation management, drought monitoring, and water resource planning in complex weather regions.</p>

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

Comparative Analysis of Deep Learning and Machine Learning Models for Evapotranspiration Prediction in Semi-Arid Regions: Statistical Model Evaluation with a Paired t-Test and Bootstrap Resampling

  • 期刊论文

  • 发表日期：

    2026年2月23日

查看全文
  • K. V. Sumith,
  • Bhavya S

摘要

Evapotranspiration is a crucial process in hydrology, agriculture, and climate studies, significantly impacting crop yield, water resources, and climate modeling, and necessitating accurate forecasting for effective management. This research evaluates the effectiveness of deep learning and machine learning techniques in forecasting evapotranspiration (ET) by utilizing climatic factors, including Rainfall, Temperature, and Sunshine hours. The Long Short-Term Memory (LSTM) model exhibited exceptional capabilities in recognizing temporal relationships and seasonal patterns within the data, achieving a test R² of 0.66, an RMSE of 0.11, and an MAE of 0.07. Statistical significance testing showed that LSTM significantly outperformed RNN (p = 0.0091), and RNN outperformed GRU (p = 0.0324), while the performance difference between LSTM and GRU was not statistically significant (p = 0.2694). The bootstrap analysis revealed low standard errors for LSTM (SE = 0.0118), RNN (SE = 0.0113), and GRU (SE = 0.0114), with 95% confidence intervals indicating reliable predictions. In contrast, traditional machine learning models like Decision Tree and GBM exhibited overfitting, with high training performance but poor generalization to unseen data. The study demonstrates the effectiveness of LSTM and RNN in predicting ET, highlighting their potential in irrigation management, drought monitoring, and water resource planning in complex weather regions.