Prediction of nitrate and sulphate dynamics in groundwater under spatiotemporal effects of urban growth using attention optimized models

Groundwater quality in urban region is increasingly at risk due to the combined effects of urban sprawl, microclimatic conditions, and large sewage generation. Domestic Reverse Osmosis (RO) systems are unable to remove nitrate and sulphate in drinking water, and leads to human health hazards. This study focuses on prediction of nitrate and sulphate levels in groundwater through integrating microclimatic conditions with urban expansion indicators. A hybrid modeling approach has been developed using an Attention-based Convolutional Neural Network (ACNN) and Bayesian Optimized Multiple Linear Regression (BO-MLR). Sentinel satellite image is used for extraction of spectral band features, with attention scores highlights the most relevant indices for groundwater contamination. The above features have been combined with field-based measurements from sprawl-affected areas in the Chengalpattu region. To refine the dataset, the FP-Growth algorithm has been applied to identify strong associations between sprawl indicators and contaminant concentrations. The BO-MLR model has achieved prediction 95% of accuracy in detection of Nitrate and Sulphate levels in drinking water, closely match to the laboratory observations. Results shows that groundwater nitrate and sulphate level increases significantly with increase in urban sprawl, with 50% increase in built-up area linked to approximately 75% higher nitrate and 60% higher sulphate levels in groundwater. The above findings highlight the urgent need for sustainable urban planning and groundwater management strategies, provides awareness and hazardous zones in Chengalpattu area.