Daily Scale Streamflow Forecasting Based-Hybrid Gradient Boosting Machine Learning Model

Abstract Hybrid model selection built with models based on machine learning (ML) and Deep learning (DL) has a significant impact on river flow predictions. Sustainable use of water resources is possible with the evaluation of basin management principles, effective natural resource management and correct water resources planning. These conditions require accurate estimation of the flows of rivers in the basin. In this study, river flow estimation was made with daily streamflow data from E12A057 (Adatepe), E12A24 (Aktaş) and E12A22 (Rüstümköy) flow measurement stations (FMSs) determined on the critical points of Sakarya Basin, which is among the important basins of Turkey. For three stations, 10 years of flow data obtained from EIEI (General Directorate of Electrical Works Survey Administration) were used. In addition, a method combining the GA-CatBoost model was proposed, which aimed to improve the performance of flow estimation. The performance of the hybrid model was compared to the CatBoost, Long-Short Term Memory (LSTM) and Linear Regression (LR) models. To analyze the performance of the model, the first 80% of the data was used for training and the remaining 20% ​​for testing the three FMS. The results revealed that the proposed hybrid model can adapt nicely with the high nonlinearity of the river flow estimation. It has been observed that the hybrid model was superior to other models in statistical measurement metrics used in the study.