Enhancing Urban Water Quality Monitoring with Remote Sensing and Deep Learning: A Scalable Approach for Real-Time Chlorophyll Prediction in Urban Lakes

Water management in regions with high concentration of population and industry has posed serious problems for maintaining water quality. Conventional approaches for water quality monitoring using grab samples with laboratory analysis are costly and insufficient to cover continuous large-scale assessments. The present study introduces Google Earth Engine and Deep Learning based prediction model that utilizes recently available satellite-based remote sensing data to predict water quality parameters such as chlorophyll-a concentration is used as an indication. The model estimates on the Modified Normalized Difference Water Index (MNDWI) and Normalized Difference Chlorophyll Index (NDCI), as principal indicators, derived from harmonised Sentinel-2 imagery over seven urban lakes. The Google Remote Information Sensing and Deep Neural Networks (GRIS-DNN) model was built with hyperparameter. A comparison with Random Forest, Support Vector Machine and other Neural Network architectures revealed that our GRIS-DNN model had a higher <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{R}^{2}$</tex> value of 0.95 and low RMSE <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$=0.89\ \mathbf{mg}/\mathbf{m}^{3}$</tex>, thus outperforming them on the prediction task for chlorophyll-a levels. The proposed method provides a rapid, robust and real-time solution for monitoring urban lake water quality management that can provide useful guidance to policy makers on best to address the challenges facing urban water resources.