An improved technique for streamflow forecasting between Turkish straits


Karsavran Y., Erdik T., Özger M.

Acta Geophysica, 2023 (SCI-Expanded) identifier

  • Publication Type: Article / Article
  • Publication Date: 2023
  • Doi Number: 10.1007/s11600-023-01216-z
  • Journal Name: Acta Geophysica
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Communication Abstracts, Compendex, Geobase, INSPEC, Metadex, Civil Engineering Abstracts
  • Keywords: Bosphorus strait, Cross-wavelet transform, Dardanelles strait, Discharge prediction, Lag time, Turkish straits system
  • Istanbul Technical University Affiliated: Yes

Abstract

The Turkish straits system (TSS), consisting of the Marmara Sea, Dardanelles and Bosphorus, transfers the Black Sea water and the Mediterranean water. To our best knowledge, finding the lag time of the Bosphorus to Dardanelles and estimating upper layer discharge of the Dardanelles using Bosphorus upper layer discharge have not been studied before in the existing literature. With this motivation, continuous wavelet transform (CWT) was applied to the upper layer discharge data of the Dardanelles and Bosphorus as a pre-processing instrument and the noise in the data was eliminated by separating them into sub-series. The wavelet coherence (WTC) and the cross-wavelet transform (XWT) were applied to detect the lag time of the Bosphorus to the Dardanelles. The upper layer flow lag time of the Bosphorus to the Dardanelles is calculated to be about 5 days. In addition, there are not sufficient studies to predict Dardanelles flow by using Bosphorus flow. In order to fill this gap in the literature, the wavelet-ANN (WANN) and stand-alone ANN were employed to estimate the upper layer discharges of the Dardanelles from Bosphorus. Performances of model were estimated by RMSE and Pearson’s R 2. ANN and WANN models were developed to predict discharges with lead times of up to 7 days (168 h). Using the wavelet transform as a pre-processing tool significantly enhanced estimations in comparison to the stand-alone ANN model. While stand-alone ANN model cannot make accurate predictions for any lead time, the WANN model can make correct forecasts of Dardanelles upper layer discharges for up to 3 days (72 h).