Remote Sensing Approaches for Water Quality Monitoring in India: A Systematic Review of Algorithms and Models

Vidya N. Kawtikwar, Prof. Varsha Bhosale

Background: Monitoring of water quality in India, specifically reservoirs and inland basins, is important for the sustainability of the environment. As the traditional in-situ methods are spatially limited, there is a need for the integration of multispectral remote sensing and artificial intelligence (AI) for the large-scale assessment. Objectives: This review analyses the transition of technologies used for monitoring of water quality from traditional models to the advanced hybrid models of deep learning (DL). Mainly, this study shows the key insight of the 2020–2025 literature by setting a baseline for retrieval techniques for water quality parameters from traditional optical parameters to advanced non-optical indicators. Method: This study systematically screened 78 research papers from the databases, viz., Web of Science, Google Scholar and Scopus, from the time frame of 2020 to 2025. The inclusion criteria targeted papers using UAV, Sentinel-2 and Landsat-8/9 UAV platforms used for water quality monitoring of Indian river basins and reservoirs, while exclusion criteria have removed the papers which were lacking in AI-based validation or those predating 2020. Keywords used are "Deep Learning”, “Remote Sensing”, "WQI", "Sentinel-2" and "Indian Inland Waters”. The study consists of a comparison of parameters such as WQI, turbidity and chlorophyll-a across models for machine learning and deep learning. The comparative study is performed by analysing the Coefficient of Determination (R²) values and is presented using a four-tier methodological framework which bridges the gap between satellite data and drone data, a parameter-sensor heatmap, and a comparative summary matrix of all 26 core references. Findings: The analysis proved that hybrid architectures, specifically CNN-LSTM models, have achieved the accuracy of R² as 0.9999 as compared to earlier support vector machine and random forest approaches. A major milestone identified is the successful use of the COVID-19 lockdown as a baseline for turbidity monitoring. Key bottlenecks include the persistent "black box" nature of deep learning and optical interference during the Indian monsoon. The roadmap indicates a shift toward Explainable AI (XAI) and the integration of SAR-optical data fusion to ensure all-weather monitoring. Significance: This review suggests new insight by focusing exclusively on the "Artificial Intelligence (AI) Remote Sensing Synergy (RS)" post-2020, which is an area which is rapidly evolving beyond the scope but was completely ignored in previous global reviews. Unlike the earlier studies which have ignored the model transparency, this review considers explainability as a core metric which provides a unique "cross-scale" perspective that shows the linking between ultra-high-resolution UAV data and satellite-based basin monitoring which is specifically tailored to the complex hydrodynamics of the Indian subcontinent. Besides, it also focuses on technological fusion that can contribute to better management of water bodies in India. Keywords: Remote Sensing, Deep Learning, Water Quality Index (WQI), Indian river basins, Sentinel-2, Explainable AI (XAI)