Coastal groundwater quality prediction using objective-weighted WQI and machine learning approach

Area of Science:

• Environmental Science
• Water Resource Management
• Artificial Intelligence

Background:

• Groundwater levels are declining in coastal zones, necessitating effective water quality monitoring.
• Accurate prediction of groundwater status is vital for sustainable water management and public health.

Purpose of the Study:

• To compare entropy and critic weight-based water quality index (WQI) methods.
• To develop and validate a multi-layer perceptron artificial neural network (MLP-ANN) model for WQI prediction.
• To identify contaminated zones using Geographic Information System (GIS).

Main Methods:

• Collected 1000 water sampling datasets from eastern India (2018-2022).
• Estimated entropy-based WQI (ENW-WQI) and critic-based WQI (CRITIC-WQI).
• Developed MLP-ANN models using different data partitioning and hidden neuron numbers, validated with correlation and trial-error analysis.
• Generated spatial distribution maps using inverse distance weighted interpolation.

Main Results:

• 65-67% of water samples were rated excellent to good for drinking.
• The CRITIC-WQI-MLP-ANN-II model achieved the highest accuracy (R²=0.986, NSE=0.98, error rate=0.49%).
• GIS-based WQI maps identified areas with varying drinking water quality.

Conclusions:

• The CRITIC-WQI-MLP-ANN-II model demonstrates superior performance for WQI prediction.
• GIS mapping effectively visualizes groundwater quality distribution.
• The findings support planning for safe drinking water provision in coastal regions.