Answers are grounded in this paper and cite their sources.
Total phosphorus (TP), a key driver of aquatic eutrophication, requires effective monitoring for water quality management. Traditional in situ methods are constrained by high costs and spatial limitations, driving the advancement of remote sensing as a complementary, synoptic tool. This review synthesizes the state-of-the-art in hyperspectral TP retrieval, focusing on inland and coastal waters. We outline the end-to-end retrieval framework, from data acquisition and preprocessing to feature engineering, model development, and spatial mapping, and systematically review hyperspectral data sources, including ground-based, autonomous aerial vehicle, airborne, and satellite platforms. The methodological evolution from empirical models to advanced machine learning and deep learning approaches is critically examined, alongside emerging hybrid and explainable artificial intelligence (AI) frameworks that integrate multisource environmental data to enhance accuracy and interpretability. We detail diverse applications across aquatic systems, from eutrophic lakes and urban rivers to estuaries and coastal bays, highlighting both regional case studies and large-scale assessments. Persistent challenges are identified, including the indirect optical nature of TP, model transferability limitations, atmospheric correction uncertainties, spectral interference from covarying constituents, and data synchronization issues. Opportunities are discussed in the context of AI advancements, multiplatform data fusion, mechanistic phosphorus fraction retrieval, and moves toward standardized global datasets. This review aims to guide researchers and practitioners toward more robust, interpretable, and scalable TP monitoring solutions, ultimately supporting improved water quality governance and ecological sustainability.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: