Integration of extreme learning machines with CEEMDAN and VMD techniques in the prediction of the multiscalar standardized runoff index and standardized precipitation evapotranspiration index

KATİPOĞLU O. M.

Natural Hazards, cilt.120, sa.1, ss.825-849, 2024 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 120 Sayı: 1
  • Basım Tarihi: 2024
  • Doi Numarası: 10.1007/s11069-023-06238-w
  • Dergi Adı: Natural Hazards
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, IBZ Online, PASCAL, Aerospace Database, Agricultural & Environmental Science Database, Aquatic Science & Fisheries Abstracts (ASFA), CAB Abstracts, Communication Abstracts, Environment Index, Geobase, INSPEC, Metadex, PAIS International, Pollution Abstracts, Veterinary Science Database, DIALNET, Civil Engineering Abstracts
  • Sayfa Sayıları: ss.825-849
  • Anahtar Kelimeler: Machine learning, Middle Euphrates basin, Signal decomposition, Standardized precipitation evapotranspiration index, Standardized runoff index
  • Erzincan Binali Yıldırım Üniversitesi Adresli: Evet

Özet

Accurate prediction of droughts is vital for effectively managing droughts, assessing drought risks and impacts, drought early warning systems, drought preparedness, and mitigation policies. This study integrated the extreme learning machines algorithm into the complementary ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and variational mode decomposition (VMD) techniques to predict 1-month lead time meteorological and hydrological droughts. Standardized precipitation evapotranspiration index (SPEI) and standardized runoff index (SRI) values were calculated as meteorological and hydrological drought indicators, respectively. The effects of the previous SRI and SPEI values were evaluated to estimate 3- and 12-month SRI-based hydrological droughts. The previous SPEI values were used to estimate SPEI values. The cross-correlation matrix and partial correlation function were used to determine the model input combinations. It is recommended to input delayed drought indices up to 3 months as input to the model for predicting droughts with a lead time of 1 month. The performance of the models was compared with statistical indicators such as coefficient of determination, mean square error, and mean absolute error, and scatter diagram and violin box plot. As a result of the analyses, it was determined that decomposition techniques improved the drought prediction accuracy of the extreme learning machines (ELM) model. The highest prediction performance (R 2: 0.926, MSE: 0.084, and MAE: 0.229) was achieved in the prediction of SRI12 (t + 1) values using the VMD-ELM hybrid approach with the input combination of SRI12 (t) and SRI12 (t − 1) values at the Hinis station. In addition, it has been revealed that the VMD decomposition method provides more successful separation than CEEMDAN in both SRI and SPEI estimations. The study outputs are essential in managing water resources, hydroelectric energy production, sizing of water structures, and transboundary waters.