Hybrid machine learning models for groundwater level prediction in a snow-dominated region: An evaluation of EEMD, VMD and EWT decomposition techniques

GEZİCİ K., KATİPOĞLU O. M., ŞENGÜL S.

Hydrological Processes, cilt.38, sa.5, 2024 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 38 Sayı: 5
  • Basım Tarihi: 2024
  • Doi Numarası: 10.1002/hyp.15169
  • Dergi Adı: Hydrological Processes
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Aerospace Database, Aqualine, Aquatic Science & Fisheries Abstracts (ASFA), Artic & Antarctic Regions, CAB Abstracts, Communication Abstracts, Compendex, Environment Index, Geobase, INSPEC, Metadex, Pollution Abstracts, Veterinary Science Database, Civil Engineering Abstracts
  • Anahtar Kelimeler: empirical wavelet transform, extreme learning machine, groundwater level, machine learning, signal decomposition, variational mode decomposition
  • Erzincan Binali Yıldırım Üniversitesi Adresli: Evet

Özet

Water scarcity is a pressing issue, intensified by factors such as population growth and industrialization. Hence, it is crucial to monitor, conserve and analyse groundwater resources, which are essential sources of clean and usable water. This study examines changes in groundwater levels (GWLs) in northeastern Turkey's mountainous and snow-covered area. The primary objective is to assess the effectiveness of integrated machine learning models, specifically, the extreme learning machine (ELM) technique combined with signal decomposition techniques such as ensemble empirical mode decomposition (EEMD), variational mode decomposition (VMD) and empirical wavelet transform (EWT) for monthly GWL prediction models. Seventy percent of the accessible data is allocated for training, while 30% is designated for testing. A correlation matrix involving precipitation, temperature, relative humidity and GWL parameters is generated with inputs that possess significant correlations being selected, such as GWLt−1, GWLt−2, RHt, RHt−1 and RHt−2. To evaluate model results, various metrics, including mean squared error, mean absolute error, mean absolute percentage error, mean bias error, bias factor, determination coefficient, Nash-Sutcliffe efficiency, as well as tools such as box plots, Taylor diagrams and radar charts, are utilized to compare outcomes during the interpretation phase. The results of the analyses show that applying data decomposition methods such as EEMD, VMD and EWT significantly improves the performance of the ELM algorithm in predicting GWLs. VMD-ELM is the most accurate for GWL forecasting among the approaches examined. The R2 values of the most successful models established in the two wells are 0.993 and 0.905. The outcomes of this research hold significance for decision-makers and policymakers as it offers informative insights into aquifer surveillance, irrigation strategizing and efficient administration of water resources.