Scholarlink

<p>Accurate precipitation estimation is essential for reliable hydrological and atmospheric modeling; however, satellite-based products often require bias correction before they can be reliably applied. Previous studies have addressed this challenge, but many suffer from incomplete error characterization, limited integration of terrain and climatic factors, and insufficient exploration of hybrid statistical and data-driven approaches. To overcome these gaps, this study proposes hybrid nested frameworks that combine collocation techniques with advanced machine learning algorithms while explicitly incorporating terrain attributes and climatic variables to improve predictive accuracy. We applied Triple Collocation (TC), Multiple Collocation (MC), and a suite of machine learning models, including Random Forest (RF), XGBoost, Support Vector Machines (SVM), Artificial Neural Networks (ANN), and Long Short-Term Memory networks (LSTM). The most accurate approach—a nested hybrid model integrating LSTM with TC (NESTED-LSTM-TC)—achieved an 87.1% bias reduction, a mean absolute error (MAE) of 7.12&#xa0;mm, a root mean square error (RMSE) of 9.10&#xa0;mm, and a Kling–Gupta Efficiency (KGE) score of 0.84. These findings demonstrate the strong potential of hybrid nested frameworks to substantially improve satellite-derived precipitation estimates by fusing statistical rigor with advanced machine learning techniques.</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

A Multi-Level Data Mining and Blending Framework for Enhanced Accuracy of Satellite-Based Datasets in Hydrological Applications

  • 期刊论文

  • 发表日期：

    2025年12月20日

查看全文
  • Mehran Besharatifar,
  • Mohammadnabi Jalali,
  • Keivan Rahsepas

摘要

Accurate precipitation estimation is essential for reliable hydrological and atmospheric modeling; however, satellite-based products often require bias correction before they can be reliably applied. Previous studies have addressed this challenge, but many suffer from incomplete error characterization, limited integration of terrain and climatic factors, and insufficient exploration of hybrid statistical and data-driven approaches. To overcome these gaps, this study proposes hybrid nested frameworks that combine collocation techniques with advanced machine learning algorithms while explicitly incorporating terrain attributes and climatic variables to improve predictive accuracy. We applied Triple Collocation (TC), Multiple Collocation (MC), and a suite of machine learning models, including Random Forest (RF), XGBoost, Support Vector Machines (SVM), Artificial Neural Networks (ANN), and Long Short-Term Memory networks (LSTM). The most accurate approach—a nested hybrid model integrating LSTM with TC (NESTED-LSTM-TC)—achieved an 87.1% bias reduction, a mean absolute error (MAE) of 7.12 mm, a root mean square error (RMSE) of 9.10 mm, and a Kling–Gupta Efficiency (KGE) score of 0.84. These findings demonstrate the strong potential of hybrid nested frameworks to substantially improve satellite-derived precipitation estimates by fusing statistical rigor with advanced machine learning techniques.