Application of soft computing to predict water quality in wetland

Area of Science:

• Environmental Science
• Water Resource Management
• Computational Intelligence

Background:

• Water quality prediction is vital for human and ecosystem health.
• Free surface wetlands require accurate water quality monitoring.
• Soft computing offers efficient methods for complex environmental predictions.

Purpose of the Study:

• To predict the Water Quality Index (WQI) in a free surface wetland.
• To compare the performance of Adaptive Neuro-Fuzzy System (ANFIS), Artificial Neural Networks (ANNs), and Group Method of Data Handling (GMDH) for WQI prediction.
• To identify key water quality parameters influencing WQI.

Main Methods:

• Monitoring of 11 water quality parameters (conductivity, SS, BOD, AN, COD, DO, temperature, pH, phosphate, nitrite, nitrate) across 17 wetland points over 14 months.
• Application of ANFIS, ANNs, and GMDH models for WQI prediction.
• Sensitivity analysis using ANFIS to determine significant WQI predictors (pH, COD, AN, SS).

Main Results:

• ANFIS achieved the highest prediction accuracy with Nash-Sutcliffe Efficiency (NSE) of 0.9634 and Mean Absolute Error (MAE) of 0.0219.
• ANNs provided comparable results (NSE = 0.9617, MAE = 0.0222).
• GMDH offered acceptable prediction accuracy (NSE = 0.9594, MAE = 0.0245) suitable for practical applications.

Conclusions:

• ANFIS is the most effective soft computing technique for WQI prediction in wetlands.
• ANNs and GMDH are viable alternatives for water quality prediction.
• These computational methods offer reduced running time and high speed, making them suitable for global aquatic environment management.