Retrieval of Chlorophyll-A Concentration via QA-Guided Adaptive Selection of Multiple Atmospheric Correction Algorithms

Atmospheric correction (AC) uncertainties critically constrain satellite chlorophyll-a (CHLA) retrieval in optically complex coastal waters. Existing AC algorithms perform divergently across water types, and no single algorithm is universally optimal. Although multi-source fusion has been widely explored, current studies predominantly integrate satellite sensors or inversion models while neglecting uncertainties inherent to the preprocessing AC step. In this study, we developed a pixel-wise AC optimization method using the QA score model to evaluate and select spectrally complementary outputs from multiple AC algorithms. Applied to GOCI data over the Shandong Peninsula, four algorithms (GDPS 1.3, GDPS 2.0, Seadas_Default, and Seadas_MUMM) were employed. For each pixel, the optimal remote sensing reflectance (Rrs) was selected based on QA scores, followed by CHLA retrieval via the YOC model. Validation against 96 in situ measurements demonstrated significantly improved accuracy (r = 0.868, RMSE = 0.582 μg/L, ε = 16.9%) compared with any single AC method. This study confirms that pixel-wise AC optimization and selection effectively suppress algorithm-specific uncertainties, providing a robust strategy for enhancing satellite-derived CHLA estimates in complex coastal waters.