Dissolved oxygen modelling of the Yamuna River using different ANFIS models

Area of Science:

• Environmental Science
• Water Quality Management
• Machine Learning Applications

Background:

• Dissolved oxygen (DO) is a critical indicator of river health and ecosystem sustainability.
• Accurate DO estimation is vital for effective water quality improvement plans and riverine ecosystem management.
• Machine learning offers powerful tools for simulating and predicting complex environmental parameters like DO.

Purpose of the Study:

• To simulate and predict dissolved oxygen (DO) levels in the Delhi stretch of the Yamuna River using machine learning techniques.
• To evaluate the performance of adaptive neuro fuzzy inference system-grid partitioning (ANFIS-GP) and subtractive clustering (ANFIS-SC) for DO prediction.
• To identify the optimal combination of physiochemical parameters for accurate DO simulation.

Main Methods:

• A 5-year dataset of physiochemical parameters from the Yamuna River was utilized.
• Four distinct models (M1-M4) were developed using varying input parameter combinations.
• Adaptive neuro fuzzy inference system-grid partitioning (ANFIS-GP) and subtractive clustering (ANFIS-SC) were employed for DO simulation.
• Model performance was assessed using root mean square error and coefficient of determination (R²).

Main Results:

• Both ANFIS-GP and ANFIS-SC models demonstrated adequate and accurate DO prediction capabilities.
• ANFIS-GP generally outperformed ANFIS-SC in simulating DO levels.
• Model M4, using a specific combination of input parameters with ANFIS-GP, achieved a high coefficient of determination (R²) of 0.953.
• ANFIS-SC with M4 yielded an R² of 0.911, indicating strong predictive performance but lower than ANFIS-GP.

Conclusions:

• Machine learning models, especially ANFIS-GP, are effective tools for simulating and predicting dissolved oxygen in river systems.
• The selection of appropriate input parameters significantly influences the accuracy of DO prediction models.
• ANFIS-GP provides a robust and accurate method for assessing and managing water quality parameters like DO in the Yamuna River.