Assessing Water Quality through Remote Sensing: A Regression-Based Approach with Sentinel-2 Data

Abstract

Monitoring water quality is crucial for both human health and environmental sustainability. Conventional monitoring techniques, which depend on laboratory analysis and point-based sampling, frequently do not cover spatial and temporal data. This study evaluated the water quality along Ahmedabad, India's Sabarmati Riverfront, via the Google Earth Engine (GEE) and remote sensing technology. Key water characteristics, such as pH, turbidity (Tur), Total Suspended Solids (TSS), Total Solids (TS), Dissolved Oxygen (DO), Biochemical Oxygen Demand (BOD), Total Phosphorus (TP), Fecal Coliform (FC), and Ammonia (NH₃), were estimated via the analysis of Sentinel-2 satellite images. To determine the relationships between in-situ water quality data and satellite-derived spectral indices, an empirical model based on regression was developed. Significant seasonal and regional changes in water quality are shown by the results, with areas showing desirable amounts of TSS, BOD, and FC. While the model exhibited modest association (R² = 0.62) with haziness contamination indicators such as turbidity, it was good at predicting the accuracy of parameters such as TP (R² = 0.75), TSS (R² = 0.76) and pH (R² = 0.80). Real-time water quality evaluation is made affordable and scalable by the combination of remote sensing and GIS, which also offers important insights for pollution prevention and resource management. To improve prediction capabilities, future research can use hyperspectral imaging and machine learning techniques to increase model accuracy.