作者：Li, N., Zhang, Y., Shi, K., Zhang, Y., Lin, W., Luo, X., Qin, B., Zhu, G. & Duan, H.

Accurate quantification of chlorophyll-a (Chla) is vital for understanding, assessing, managing eutrophication, aquatic ecosystem health, and biogeochemical cycling. However, satellite-based Chla estimation in Case II waters remains challenging due to complex optical properties, limited synchronized field data, and uncertainties in spectral-temporal matching. This study introduces a novel approach that integrates Sentinel-2 multispectral imagery with coincident hyperspectral proximal sensing (HPS) measurements to enhance the Chla estimation accuracy in optically complex waters. First, a band-conversion fusion method with high consistency (slope = 1 and R-2 = 0:95) was proposed by leveraging strong correlations between Sentinel-2 and HPS spectra. Subsequently, a high-precision machine learning (ML)-based fused Chla model [R-2 = 0:93, normalized root-mean-squared error (NRMSE) = 24.2%, and mean absolute percentage error (MAPE) = 31.6%] was developed using a fusion dataset combining HPS measurements (N = 1728) and Sentinel-2 data (N = 100). This model significantly outperformed the unfused Sentinel-2-only model and four existing literature algorithms. When applied to Sentinel-2 time-series data (2016-2024) over Lake Taihu, the fused model revealed a notable eutrophication mitigation trend, with Chla decreasing from 30.78 to 19.56 mu g /L. Comparative analyses demonstrated that fusing HPS data improved Sentinel-2-derived Chla estimation performance because strict temporal synchronization of HPS measurements reduced spectral uncertainty and enhanced sensitivity to Chla variability. Recalibrating Sentinel-2 spectra reduced MAPE from 41% to 9.7%, and expanding field-matched datasets enhanced the model reliability and robustness by 23.4% and 10.8%, respectively. The study underscores the transformative potential of multisource data fusion for advancing satellite-based water-quality monitoring, particularly in optically complex aquatic systems. The proposed framework o ffers a transferable methodology for estimating other biogeochemical parameters through integrated spaceborne and proximal sensing approaches, providing a foundation for improved environmental monitoring and management.