Retrieving Inland Water Quality Parameters via Satellite Remote Sensing: Sensor Evaluation, Atmospheric Correction, and Machine Learning Approaches

Mohsen Ansari, Anders Knudby, Meisam Amani, Michael Sawada

Satellite remote sensing provides a cost-effective and large-scale alternative to traditional methods for retrieving water quality parameters for inland waters. Effective water quality parameter retrieval via optical satellite remote sensing requires three key components: (1) a sensor whose measurements are sensitive to variations in water quality; (2) accurate atmospheric correction to eliminate the effect of absorption and scattering in the atmosphere and retrieve the water-leaving radiance/reflectance; and (3) a bio-optical model used to estimate water quality from the optical signal. This study provides a literature review and an evaluation of these three components. First, a review of decommissioned, active, and upcoming satellite sensors is presented, highlighting their advantages and limitations, and a ranking method is introduced to assess their suitability for retrieving chlorophyll-a, colored dissolved organic matter, and non-algal particles in inland waters. This ranking can aid in selecting appropriate sensors for future studies. Second, the strengths and weaknesses of atmospheric correction algorithms used over inland waters are examined. The results show that no atmospheric correction algorithm performed consistently across all conditions. However, understanding their strengths and weaknesses allows users to select the most suitable algorithm for a specific use case. Third, the challenges, limitations, and recent advances of machine learning use in bio-optical models for inland water quality parameter retrieval are discussed. Machine learning models have limitations, including low generalizability, low dimensionality, spatial/temporal autocorrelation, and information leakage. These issues highlight the importance of locally trained models, rigorous cross-validation methods, and integrating auxiliary data to enhance dimensionality. Finally, recommendations for promising research directions are provided.