Data-driven coastal nutrients estimation: current status, research gaps, and future developments

Driven by warming and anthropogenic impacts, global coastal waters suffer from the escalating eutrophication crisis, making it essential to understand the dynamics of nutrient discharge and enrichment. While satellites can retrieve surface water optical properties, the estimation of coastal nutrients from space remains a long-standing challenge due to their non-optical properties. Recently, artificial intelligence, particularly machine learning, has enabled the spatiotemporal estimation of surface nutrients from satellite observations, which represents a potential breakthrough for sustainable water management. In this review, we summarize the mechanisms governing coastal biogeochemical cycling and identify the estimation complexity that international scholars have been striving to solve over the past decade. We demonstrate that data-driven methods have shown preliminary success in empowering remote sensing retrieval of nitrogen, phosphorus, and silicon. Despite these advances, we note that significant challenges remain for effective monitoring of coastal eutrophication due to the limited interpretability of data-driven estimation and data scarcity. A thorough analysis is provided on the developments and limitations of nutrient retrieval, the potential of integrating data-driven methods with marine biogeochemical mechanisms, and future opportunities to enhance marine ecosystem dynamic models. Ultimately, our aim is to advance satellite observations and data-driven estimation at the forefront of environmental science to develop the precise monitoring, assessment, and protection of coastal waters.