Study on the reconstruction of daily-scale dynamics of chlorophyll concentration in the South China Sea

The South China Sea (SCS) exhibits complex bio-geochemical processes, characterized by significant spatial heterogeneity and temporal variability in surface chlorophyll-a(Chl-a) concentrations. However, satellite data for this region are severely impacted by cloud cover, which hinders a systematic understanding of the spatiotemporal distribution characteristics of Chl-a. This study first employs the Data-Interpolating Convolutional Auto-Encoder (DINCAE) to reconstruct daily remotely sensed Chl-a concentration data from 2003 to 2023 in the central and northern parts of the SCS. Subsequently, Empirical Orthogonal Function (EOF) decomposition is applied to systematically analyze the spatiotemporal evolution characteris-tics of Chl-a concentrations. In the validation dataset, the DINCAE reconstruction results achieved a coefficient of determination (R2 ) of 0.98 and a root mean square error (RMSE) of 0.043 mg·m⁻³. The EOF analysis revealed that the primary spatial mode aligns well with the winter Chl-a distribution patterns in this region, particularly highlighting the "inverted V" and "two-wing" morphology near the Luzon Strait during winter. This finding also indirectly validates the effectiveness of the reconstruction method, and the main temporal mode indicates peak values occurring during the winter months.