Merging bio-optical products from multiple ocean color missions requires consistent remote sensing reflectance (R_rs) measurements. However, differences in sensor characteristics and data processing have led to significant discrepancies in the R_rs products, particularly in polar oceans. In this study, we introduce a Cross-Sensor Atmospheric Correction (CSAC) algorithm that uses neural networks to achieve consistent R_rs retrievals between Visible Infrared Imaging Radiometer Suite (VIIRS) and Moderate Resolution Imaging Spectroradiometer (MODIS)-Aqua in the Ross Sea, Southern Ocean. The CSAC converts VIIRS top-of-atmosphere reflectance (ρ_t-VIIRS) into R_rs at equivalent MODIS-Aqua bands (R_rs-MAE), where a large dataset of matched high-quality MODIS-Aqua R_rs and ρ_t-VIIRS was used to train the algorithm (N > 9 million). With CSAC, the averaged absolute percent difference (APD) between VIIRS-coverted R_rs-MAE and the standard MODIS-Aqua R_rs (R_rs-MA) ranges from 8.8% to 22.3% across wavelengths of 412–667, which are significantly better than the APD of standard VIIRS R_rs product (R_rs-VIIRS) (14.7%–85.9%). Importantly, CSAC also increases the number of valid R_rs retrievals in the Ross Sea compared to R_rs-VIIRS, greatly enhancing data coverage, particularly for observation with high solar and viewing angles. These results highlight the effectiveness of CSAC in generating higher-quality, more consistent ocean color data across sensors, which is essential for long-term monitoring of polar marine ecosystems.

cross-sensor atmospheric correction, remote sensing reflectance, neural networks, Ross Sea, ocean color