COMPARISON OF DIFFERENT ANN (FFBP GRNN F) ALGORITHMS AND MULTIPLE LINEAR REGRESSION FOR DAILY STREAMFLOW PREDICTION IN KOCASU RIVER-TURKEY

Burgan H. İ.

FRESENIUS ENVIRONMENTAL BULLETIN, cilt.31, sa.5, ss.4699-4708, 2022 (SCI-Expanded)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 31 Sayı: 5
  • Basım Tarihi: 2022
  • Dergi Adı: FRESENIUS ENVIRONMENTAL BULLETIN
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Aerospace Database, Aqualine, Aquatic Science & Fisheries Abstracts (ASFA), CAB Abstracts, Chemical Abstracts Core, Communication Abstracts, Environment Index, Geobase, Greenfile, Metadex, Pollution Abstracts, Veterinary Science Database, Civil Engineering Abstracts
  • Sayfa Sayıları: ss.4699-4708
  • Akdeniz Üniversitesi Adresli: Evet

Özet

Flow gauging stations in hydrological basins are mostly installed on main rivers. It is known that how difficult the prediction of daily flow with its stochastic and complicated structure. In this study, the application of time-lagged streamflow records of a gauging station is proposed as an accurate method for flow prediction. In this way, five time-lagged scenario is evaluated as daily streamflow prediction function station at Kocasu River, Turkey. At the first stage, typical three-layer feed forward back propagation (FFBP) neural networks is applied as an ANN method to reach the best time-lagged solution for the river. Additionally, other ANN algorithms as generalized regression neural networks (GRNN) and radial basis function (RBF) neural networks and also multiple linear regression (MLR) method are applied in order to comparison of the mentioned ANN and MLR techniques. Root mean square error (RMSE) and determination coefficients (R2) are calculated to evaluate the performance of the techniques. According to the results, daily records of the station are sufficient to achieve high efficiency value which can be proposed as the most reasonable daily streamflow prediction model for Kocasu river, which is in the southern part of the Marmara Sea. The performance of FFBP algorithm for daily flow prediction studies is the best one in all other techniques. At the same time, these ANN algorithms can be used not only in flow prediction, but also they can be used for the purpose of water resources management in hydrological basins by estimating extreme events as floods and droughts.