Hybrid model for daily streamflow and phosphorus load prediction.

Area of Science:

• Environmental Science
• Hydrology
• Water Resource Management

Background:

• Environmental factors like climate and land use changes significantly impact water quality.
• Existing predictive models (process-based and data-driven) have limitations in accuracy and scope.
• Accurate prediction of water quality parameters is crucial for effective watershed management.

Purpose of the Study:

• To develop and evaluate a novel hybrid model, SWAT-RETAIN, for predicting daily streamflow and total phosphorus (TP) load.
• To improve the accuracy and temporal resolution of watershed water quality predictions.
• To address the limitations of existing individual modeling approaches.

Main Methods:

• Developed a hybrid model integrating the Soil and Water Assessment Tool (SWAT) with the Reverse Time Attention Mechanism (RETAIN).
• Utilized spatial, meteorological, and hydrological data to build the SWAT model for high temporal resolution data generation.
• Applied the SWAT-RETAIN model to the Hwangryong River watershed in South Korea for validation.

Main Results:

• The SWAT-RETAIN model demonstrated high accuracy in predicting streamflow (NSE: 0.45) and TP load (NSE: 0.50) on test sets.
• Performance was validated using flow duration and load duration curves.
• The model achieved effective simultaneous prediction of multiple water quality variables.

Conclusions:

• The SWAT-RETAIN hybrid model offers enhanced temporal resolution and predictive performance for watershed water quality.
• This integrated approach effectively overcomes limitations of standalone models.
• SWAT-RETAIN shows potential for application in predicting various water quality variables in larger and more complex watersheds.