Sarıgöl, Metin; Katipoğlu, Okan Mert; Aktürk, Gaye

Flood routing is essential for water resources management and flood mitigation, yet traditional methods often struggle to represent complex hydraulic processes and rapidly changing flow conditions. In this study, hydrological models (Muskingum and SCS), hydraulic methods (Kinematic Wave, Muskingum–Cunge, and Dynamic Wave), and deep learning (DL) algorithms (Autoencoder, Deep Neural Network (DNN), Gated Recurrent Unit (GRU), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Recurrent Neural Network (RNN)) were applied to predict flood routing. Model performance was evaluated using several statistical indicators, including RMSE, AIC, MAE, KGE, NSE, R², Pbias, and MBE. The results showed that the Autoencoder model provided the best predictive performance (RMSE: 0.30, MAE: 0.17, NSE: 0.97, R²: 0.99), while the RNN model produced the weakest results. Other DL models, particularly DNN, CNN, and LSTM, also demonstrated strong predictive capability. Among the hydraulic approaches, the Kinematic Wave method yielded the most accurate results, whereas the SCS model showed the best performance among hydrological routing methods. Overall, the findings indicate that DL models generally outperform classical routing approaches in flood prediction, with the Autoencoder architecture emerging as the most effective model under the conditions of the present study.

Katipoğlu, Okan Mert; Kartal, Veysi; Tuğrul, Türker; Yavuz, Veysel Süleyman; Ariman, Sema

This study developed an innovative hybrid modelling approach to predict one-month-ahead water levels (LWL (T + 1)) for two major lakes in Türkiye (Eber and Eğirdir), aiming to model complex dynamics of hydrological systems and optimize sustainable water resources management. In the proposed methodological framework, Singular Spectrum Analysis (SSA) was employed to decompose hydrometeorological time series into sub-components, while Mutual Information (MI) and Partial Autocorrelation Function (PACF) techniques were utilized to determine optimal input variables. The hybrid model architecture, incorporating Deep Neural Networks (DNN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and CatBoost algorithms, was optimized using 70% of the data for training and 30% for testing, and its performance was evaluated through deterministic and probabilistic metrics. The analyses revealed that SSA-based data preprocessing significantly improved the performance of prediction models, and depending on lake characteristics, CNN-based models demonstrated superior performance for Lake Eber, while DNN architecture excelled for Lake Eğirdir. The developed hybrid approach presents an original methodology that can be integrated into decision support systems for water resources management, contributing to the development of sustainable water management strategies and adaptability to similar hydrological systems.

Çırağ, Burak; Acar, Reşat; Katipoğlu, Okan Mert; Yağanoğlu, Mete; Şengül, Selim

Accurate streamflow estimation is pivotal for hydrological research and water resource management, particularly in mountainous, snow-dominated regions where missing data can significantly impair watershed modeling and water budget evaluations. This study presents an advanced hybrid machine learning framework to impute missing streamflow records at a strategically located observation station within a snow-abundant sub-basin of the Euphrates-Karasu Basin in Erzurum Province, Eastern Turkey. Leveraging daily meteorological and hydrological data collected over four decades (1983–2023), we developed and rigorously evaluated several predictive models using comprehensive performance metrics and rank-based analysis. In addition to conventional approaches-including a single Artificial Neural Network (ANN), CatBoost, and Deep Neural Network (DNN)-the study also compares innovative hybrid models in which ANN parameters were optimized using the Whale Optimization Algorithm (WOA) and the Hydrologic Cycle Algorithm (HCA). Notably, the hybrid models demonstrated superior predictive accuracy over traditional methods, thereby underscoring their potential to bridge data gaps in challenging terrains. The results affirm the efficacy of the proposed framework and its promise as a robust methodological foundation for future investigations in watershed modeling, drought assessment, and water budget analysis.