Rainfall-Runoff modelling using SWAT and eight artificial intelligence models in the Murredu Watershed, India

Area of Science:

• Hydrology and Water Resources Engineering
• Environmental Science
• Artificial Intelligence in Environmental Modeling

Background:

• Increasing global water demand necessitates precise streamflow estimation for effective watershed management.
• Rainfall-runoff models are crucial tools for understanding and predicting hydrological processes.
• The Murredu River basin faces challenges in water resource management due to growing demands.

Purpose of the Study:

• To accurately estimate monthly streamflow in the Murredu River basin using various hydrological models.
• To compare the performance of the Soil and Water Assessment Tool (SWAT) model against multiple artificial intelligence (AI) models.
• To identify the most effective model for rainfall-runoff simulation to support sustainable water resource planning.

Main Methods:

• Employed the Soil and Water Assessment Tool (SWAT) model for hydrological simulation.
• Utilized eight artificial intelligence (AI) models: k-nearest neighbour, support vector regression, linear regression, artificial neural networks, random forest, XGBoost, Histogram-based Gradient Boost, and Long Short-Term Memory (LSTM).
• Calibrated and validated models using monthly streamflow data from the Murredu River basin (1999-2005).

Main Results:

• All nine models demonstrated suitability for simulating the rainfall-runoff process.
• The Long Short-Term Memory (LSTM) model exhibited superior performance, achieving high coefficients of determination (R² = 0.97) and Nash-Sutcliffe efficiency (NSE = 0.96 in calibration; R² = 0.97, NSE = 0.92 in validation).
• Other AI and SWAT models provided satisfactory results but were less accurate than LSTM.

Conclusions:

• The Long Short-Term Memory (LSTM) model is highly effective for accurate monthly streamflow modeling in the Murredu River basin.
• Selecting advanced AI models like LSTM can significantly enhance the precision of rainfall-runoff simulations.
• Findings support the use of optimal models for improved water resource management and sustainable planning in river basins.