Scholarlink

<p>Remotely sensed data&#xa0;fused with data-driven technologies prove crucial for determining lake evaporation and effectively regulating&#xa0;reservoirs in areas with inadequate information. This study evaluates how well solar radiation (<i>Ra)</i> and temperature data from reanalysis/&#xa0;satellite measurements perform to anticipate lake evaporation on a daily and seasonal&#xa0;basis. The approach evaluates the effectiveness of five Machine Learning algorithms under three input scenarios at three meteorological stations in the Awash basin, Ethiopia: Gradient Boosting (GB), Random forest (RF), Extreme Gradient Boosting (XGBoost), Support Vector Regression (SVR), and multilayer perceptron (MLP). XGBoost performs exceptionally well on the training set in each of the three scenarios. GB and MLP performs well throughout the test set for all three&#xa0;input scenarios. Subsequently, when employing models in the Metehara, Melkasa, and Dubti stations, <i>IS1</i> is the best scenario for RMSE, NSE, and KGE. Consequently, the model predicts encouraging outcomes for each location over the Tseday and Kiremit seasons. The outcomes of the study suggest that identifying the best representative satellite/reanalysis data of temperature and solar radiation for particular areas can lead to effective performance. This investigation provides insight for simulating a reservoir's existing operating system.</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

Estimating reservoir evaporation rate: fusing remotely sensed solar radiation and temperature features with machine learning algorithms

  • 期刊论文

  • 发表日期：

    2025年2月25日

查看全文
  • Kidist Demessie Eshetu,
  • Tesema Kebede Seifu,
  • Tena Alamirew,
  • Tekalegn Ayele Woldesenbet

摘要

Remotely sensed data fused with data-driven technologies prove crucial for determining lake evaporation and effectively regulating reservoirs in areas with inadequate information. This study evaluates how well solar radiation (Ra) and temperature data from reanalysis/ satellite measurements perform to anticipate lake evaporation on a daily and seasonal basis. The approach evaluates the effectiveness of five Machine Learning algorithms under three input scenarios at three meteorological stations in the Awash basin, Ethiopia: Gradient Boosting (GB), Random forest (RF), Extreme Gradient Boosting (XGBoost), Support Vector Regression (SVR), and multilayer perceptron (MLP). XGBoost performs exceptionally well on the training set in each of the three scenarios. GB and MLP performs well throughout the test set for all three input scenarios. Subsequently, when employing models in the Metehara, Melkasa, and Dubti stations, IS1 is the best scenario for RMSE, NSE, and KGE. Consequently, the model predicts encouraging outcomes for each location over the Tseday and Kiremit seasons. The outcomes of the study suggest that identifying the best representative satellite/reanalysis data of temperature and solar radiation for particular areas can lead to effective performance. This investigation provides insight for simulating a reservoir's existing operating system.