Developement of hybrid hydrological models for the sustainable water resources management

Factors such as climate change, rapid population growth, urbanization and industrialization are increasing pressure on water resources day by day. For this reason, sustainable water resource management has become a critical issue. In this context, reliable modeling and prediction of hydrological processes play a fundamental role in the development of effective water management strategies. However, since each existing hydrological model has its own advantages and limitations, it may not be able to fully represent all hydrological processes on its own. For this reason, hybrid modelling approaches that combine the strengths of physically based and data-driven models have become increasingly important in recent years. In this study, two hydrological modeling problems were addressed in line with the goal of sustainable water resource management. In the first problem, the conceptual Technische Universität Wien (TUW) hydrological model was integrated with Support Vector Regression (SVR), a machine learning method capable of representing nonlinear relationships with high accuracy, to develop the TUW-SVR hybrid hydrological model. During the model parameter calibration phase, a new single-objective hybrid optimization approach (HS-NMS) has been proposed, combining the heuristic Harmony Search (HS) optimization technique with the Nelder-Mead Simplex (NMS) method. The developed fully hybrid simulation-optimization approach has been applied to the Gediz River Basin and its performance evaluated by comparing it with flow observation data. The results demonstrate that the TUW-SVR model and the HS-NMS optimization technique significantly improve flow predictions. In the second problem, the developed hybrid TUW-SVR model has been expanded to perform simultaneous simulation of multi-objective hydrological processes. In this context, a multi-objective calibration structure has been developed that considers not only flow but also hydrological components such as soil moisture and snow cover. In this process, single-objective Shuffled Ant Lion Optimization (SHALO) has been developed based on a single-parameter structure due to its lower computational costs. The SHALO model was subsequently expanded to produce Pareto-optimal solutions, leading to the proposal of a Multi-Objective Shuffled Ant Lion Optimization (MOSHALO). To ensure continuity in the model input data, ERA5-Land reanalysis precipitation and temperature data, ASCAT satellite soil moisture, and MODIS snow cover products were used instead of MGİ data. The developed MOSHALO optimization algorithm and hybrid TUW-SVR model were applied in the Aras Basin and Çoruh Basin, yielding successful results in the simultaneous simulation of three hydrological components. Three main contributions are offered by the results obtained in this study. First, the hybrid TUW-SVR model, which was developed by integrating physical and artificial intelligence-based models, improves the accuracy of hydrological predictions. Second, the developed HS-NMS and MOSHALO optimization algorithms provide flexible, adaptable, and computationally efficient alternatives for both single- and multi-objective calibration problems. Finally, the integration of the hybrid TUW-SVR model with the MOSHALO approach enables the high-accuracy modeling of not only flow but also hydrological variables such as soil moisture and snow cover at the watershed scale.