Prediction of maximum scour depth in river bends by the Stacking model DOI Creative Commons
Junfeng Chen,

Zhou Xiao-quan,

Lirong Xiao

и другие.

Journal of Hydroinformatics, Год журнала: 2023, Номер 25(6), С. 2625 - 2642

Опубликована: Ноя. 1, 2023

Abstract The accurate prediction of maximum erosion depth in riverbeds is crucial for early protection bank slopes. In this study, K-means clustering analysis was used outlier identification and feature selection, resulting Plan 1 with six influential features. 2 included features selected by existing methods. Regression models were built using Support Vector Regression, Random Forest (RF Regression), eXtreme Gradient Boosting on sample data from 2. To enhance accuracy, a Stacking method feed-forward neural network introduced as the meta-learner. Model performance evaluated root mean squared error, absolute percentage R2 coefficients. results demonstrate that three outperformed 2, improvements values 0.0025, 0.0423, 0.0205, respectively. Among regression 1, RF performs best an value 0.9149 but still lower than 0.9389 achieved fusion model. Compared to formulas, model exhibits superior predictive performance. This study verifies effectiveness combining analysis, predicting scour bends, providing novel approach design.

Язык: Английский

A pioneering approach to deterministic rainfall forecasting for wet period in the Northern Territory of Australia using machine learning DOI Creative Commons
Rashid Farooq, Monzur Alam Imteaz,

Fatemeh Mekanik

и другие.

Earth Science Informatics, Год журнала: 2025, Номер 18(2)

Опубликована: Фев. 1, 2025

Язык: Английский

Процитировано

0
Extreme learning machine coupled with Heuristic algorithms for daily streamflow modeling at Lake Ziway Watershed, Ethiopia DOI
Gebre Gelete, Hüseyin Gökçekuş, Zaher Mundher Yaseen‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬‬

и другие.

Journal of Hydrology, Год журнала: 2025, Номер unknown, С. 133345 - 133345

Опубликована: Апрель 1, 2025

Язык: Английский

Процитировано

0
Deep learning reveals future streamflow characteristics change and climate sensitivity DOI

Subharthi Sarkar,

Mohd Imran Khan, Rajib Maity

и другие.

Journal of Hydrology, Год журнала: 2025, Номер 660, С. 133457 - 133457

Опубликована: Май 5, 2025

Язык: Английский

Процитировано

0
Short-term Prediction Method of Reservoir Downstream Water Level Under Complicated Hydraulic Influence DOI
Jingwei Huang,

Hui Qin,

Yongchuan Zhang

и другие.

Water Resources Management, Год журнала: 2023, Номер 37(11), С. 4475 - 4490

Опубликована: Авг. 11, 2023

Язык: Английский

Процитировано

6
Application of data-driven models to predict the dimensions of flow separation zone DOI Open Access
Amin Gharehbaghi, Redvan Ghasemlounıa, Sarmad Dashti Latif

и другие.

Environmental Science and Pollution Research, Год журнала: 2023, Номер 30(24), С. 65572 - 65586

Опубликована: Апрель 22, 2023

Язык: Английский

Процитировано

4
Machine Learning Approaches for Streamflow Modeling in the Godavari Basin with CMIP6 Dataset DOI Open Access
Subbarayan Saravanan, Nagireddy Masthan Reddy, Quoc Bao Pham

и другие.

Sustainability, Год журнала: 2023, Номер 15(16), С. 12295 - 12295

Опубликована: Авг. 11, 2023

Accurate streamflow modeling is crucial for effective water resource management. This study used five machine learning models (support vector regressor (SVR), random forest (RF), M5-pruned model (M5P), multilayer perceptron (MLP), and linear regression (LR)) to simulate one-day-ahead in the Pranhita subbasin (Godavari basin), India, from 1993 2014. Input parameters were selected using correlation pairwise attribution evaluation methods, incorporating a two-day lag of streamflow, maximum minimum temperatures, various precipitation datasets (including Indian Meteorological Department (IMD), EC-Earth3, EC-Earth3-Veg, MIROC6, MRI-ESM2-0, GFDL-ESM4). Bias-corrected Coupled Model Intercomparison Project Phase 6 (CMIP6) utilized process. performance was evaluated Pearson (R), Nash–Sutcliffe efficiency (NSE), root mean square error (RMSE), coefficient determination (R2). IMD outperformed all CMIP6 modeling, while RF demonstrated best among developed both datasets. During training phase, exhibited NSE, R, R2, RMSE values 0.95, 0.979, 0.937, 30.805 m3/s, respectively, gridded as input. In testing corresponding 0.681, 0.91, 0.828, 41.237 m3/s. The results highlight significance advanced applications, providing valuable insights management decision making.

Язык: Английский

Процитировано

4
Scalable and Interpretable Forecasting of Hydrological Time Series Based on Variational Gaussian Processes DOI Open Access
Julián David Pastrana-Cortés, Julian Gil-González, Andrés Marino Álvarez-Meza

и другие.

Water, Год журнала: 2024, Номер 16(14), С. 2006 - 2006

Опубликована: Июль 15, 2024

Accurate streamflow forecasting is crucial for effectively managing water resources, particularly in countries like Colombia, where hydroelectric power generation significantly contributes to the national energy grid. Although highly interpretable, traditional deterministic, physically-driven models often suffer from complexity and require extensive parameterization. Data-driven Linear Autoregressive (LAR) Long Short-Term Memory (LSTM) networks offer simplicity performance but cannot quantify uncertainty. This work introduces Sparse Variational Gaussian Processes (SVGPs) contributions. The proposed SVGP model reduces computational compared Processes, making it scalable large datasets. methodology employs optimal hyperparameters shared inducing points capture short-term long-term relationships among reservoirs. Training, validation, analysis of approach consider dataset 23 geographically dispersed reservoirs recorded during twelve years Colombia. Performance assessment reveals that proposal outperforms baseline three key aspects: adaptability changing dynamics, provision informative confidence intervals through Bayesian inference, enhanced accuracy. Therefore, SVGP-based offers a interpretable solution multi-output forecasting, thereby contributing more effective resource management planning.

Язык: Английский

Процитировано

0
Scalable and Interpretable Forecasting of Hydrological Time-Series based on Variational Gaussian Processes DOI Open Access
Julián David Pastrana-Cortés, Julian Gil-González, Andrés Marino Álvarez-Meza

и другие.

Опубликована: Апрель 23, 2024

Accurate streamflow forecasting is crucial for effectively managing water resources, particularly in countries like Colombia, where hydroelectric power generation significantly contributes to the national energy grid. Although highly interpretable, traditional deterministic, physically-driven models often suffer from complexity and require extensive parameterization. Data-driven Linear Autoregressive (LAR) Long Short-Term Memory (LSTM) networks offer simplicity performance but cannot quantify uncertainty. This work introduces Sparse Variational Gaussian Processes (SVGPs) contributions. The proposed SVGP model reduces computational compared Processes, making it scalable large datasets. methodology employs optimal hyperparameters shared inducing points capture short-term long-term relationships among reservoirs. Training, validation, analysis of approach consider dataset 23 geographically dispersed reservoirs recorded during twelve years Colombia. Performance assessment reveals that proposal outperforms baseline three key aspects: adaptability changing dynamics, provision informative confidence intervals through Bayesian inference, enhanced accuracy. Therefore, SVGP-based offers a interpretable solution multi-output forecasting, thereby contributing more effective resource management planning.

Язык: Английский

Процитировано

0
Prediction of maximum scour depth in river bends by the Stacking model DOI Creative Commons
Junfeng Chen,

Zhou Xiao-quan,

Lirong Xiao

и другие.

Journal of Hydroinformatics, Год журнала: 2023, Номер 25(6), С. 2625 - 2642

Опубликована: Ноя. 1, 2023

Abstract The accurate prediction of maximum erosion depth in riverbeds is crucial for early protection bank slopes. In this study, K-means clustering analysis was used outlier identification and feature selection, resulting Plan 1 with six influential features. 2 included features selected by existing methods. Regression models were built using Support Vector Regression, Random Forest (RF Regression), eXtreme Gradient Boosting on sample data from 2. To enhance accuracy, a Stacking method feed-forward neural network introduced as the meta-learner. Model performance evaluated root mean squared error, absolute percentage R2 coefficients. results demonstrate that three outperformed 2, improvements values 0.0025, 0.0423, 0.0205, respectively. Among regression 1, RF performs best an value 0.9149 but still lower than 0.9389 achieved fusion model. Compared to formulas, model exhibits superior predictive performance. This study verifies effectiveness combining analysis, predicting scour bends, providing novel approach design.

Язык: Английский

Процитировано

0