Physically based vs. data-driven models for streamflow and reservoir volume prediction at a data-scarce semi-arid...

Area of Science:

• Hydrology
• Environmental Modeling

Background:

• Physically based and data-driven models are used for hydrological predictions.
• Limited comparative studies exist for data-scarce, semi-arid basins with altered hydrology.

Purpose of the Study:

• To compare the performance of a physically based model (SWAT) and a data-driven model (NARX) for reservoir volume and streamflow prediction.
• To evaluate these models in the data-scarce, semi-arid Tersakan Basin, Türkiye, which has altered hydrological regimes due to reservoirs.

Main Methods:

• Employed the Soil and Water Assessment Tool (SWAT) as the physically based model.
• Utilized the Nonlinear AutoRegressive eXogenous (NARX) model as the data-driven approach.
• Calibrated and validated both models for streamflow and reservoir volumes.

Main Results:

• The NARX model demonstrated superior performance in predicting reservoir volumes (Ladik and Yedikir) and basin outlet streamflow compared to SWAT.
• Both models performed best for Ladik reservoir volume prediction and second best for Yedikir reservoir volume prediction.
• Model performance was lowest for streamflow prediction at the basin outlet.

Conclusions:

• Data-driven models, like NARX, offer a viable alternative to physically based models, especially in data-scarce environments.
• Uncertainties in input data for physically based models, such as SWAT, may contribute to lower predictive performance.
• Comparative studies are crucial for understanding model applicability in complex hydrological settings.