Application of machine learning techniques with empirical wavelet transform and variational mode decomposition for sediment concentration estimation

ÇITAKOĞLU H., Simsek O., Spor P., Gun Y. E.

Physics and Chemistry of the Earth, cilt.141, 2025 (SCI-Expanded, Scopus) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 141
  • Basım Tarihi: 2025
  • Doi Numarası: 10.1016/j.pce.2025.104136
  • Dergi Adı: Physics and Chemistry of the Earth
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Aquatic Science & Fisheries Abstracts (ASFA), Artic & Antarctic Regions, Chimica, Communication Abstracts, Compendex, Geobase, INSPEC, Metadex, Pollution Abstracts, Civil Engineering Abstracts
  • Anahtar Kelimeler: Euphrates basin, Machine learning, Preprocessing, Sediment concentration
  • Erciyes Üniversitesi Adresli: Evet

Özet

This study compares the performances of Least-Square Support Vector Regression (LS-SVM), Generalized Additive Model (GAM), Multivariate Adaptive Regression Splines (MARS), Gaussian Process Regression (GPR) methods supported by Empirical Wavelet Transform (EWT) and Variational Mode Decomposition (VMD) preprocessing techniques for estimating sediment concentration at river stations in the Euphrates Basin. In the study, the best performing models are the VMD-GPR (R2 = 0.994, RRMSE = 12.70; NSE = 0.993) for 2102 (Murat River-Palu); 2115 (Göksu River-Malpınar) for VMD-MARS (R2 = 0.968, RRMSE = 45.99, NSE = 0.961); 2119 (Euphrates River-Kemah Strait) for EWT-MARS method (R2 = 0.998, RRMSE = 6.16, NSE = 0.998); 2133 (Munzur Stream-Melekbahçe) for VMD-GPR (R2 = 0.976, RRMSE = 28.67, NSE = 0.973); 2164 (Göynük Stream-Çayağzı) for VMD-GPR (R2 = 0.982, RRMSE = 25.86, NSE = 0.980); For 2166 (Peri Suyu-Loğmar) EWT-MARS (R2 = 0.993, RRMSE = 13.83, NSE = 0.993), for 2176 (Tacik Deresi-Mutu) EWT-GPR (R2 = 0.990, RRMSE = 17.48). According to performance criteria, VMD and EWT were more compatible with GPR and MARS machine learning methods. These results reveal that machine learning and artificial intelligence techniques offer a strong alternative for sediment transport estimation. It was observed that the VMD-GPR (three stations) and EWT-MARS models stood out with their low error rates and high accuracy levels. This study's findings provide critical data regarding dam management, hydraulic structures engineering, and basin-based erosion analyses, and they significantly contribute to environmental and hydrological modelling studies.