Akademik Veri Yönetim Sistemi
International Journal of Sediment Research, 2025 (SCI-Expanded, Scopus)
In studies on water resources planning and management, regular and complete hydrological data such as streamflow and sediment data are needed. Since the existing data generally do not fully reflect the entire process, the process needs to be modeled in order to make more reliable decisions. The aim of this study is to investigate the possibilities of estimating the sediment amount with the ANN technique, which can be used in many areas today, with the streamflow and sediment measured from 20 sediment gauging stations (SGS) established by State Hydraulic Works (SHW) in the Euphrates-Tigris Basin and to try to determine the most appropriate network structure. The ANN structures to be used were determined as the most commonly used Radial Basis Artificial Neural Network (RBANN), Feed Forward Back Propagation Artificial Neural Network (FFBP) and Multilayer Artificial Neural Network (MLP). The obtained results were compared with the Multiple Linear Regression (MLR) method. The highest R2 values obtained were determined as 0.9683 and 0.9969 in the RBANN model, 0.9546 and 0.9820 in the MLP model, 0.9735 and 0.9732 in the FFBG model with the CG and LM algorithms, respectively. When only the mean values of the test values according to the ANN models were examined, the highest value was again obtained as 0.8507 in the RBANN and LM algorithms. In terms of sediment estimation, the highest R2 value in the ANN analysis was found in the RBANN model LM algorithm as 0.9804 in the train phase, 0.9969 in the testing phase and 0.9970 in the cross-validation phase.