Purpose. This article discusses the use of the blue colour index in atmospheric correction for remote sensing instruments. The blue colour index is defined as the ratio of remote sensing reflectance in two shortwave channels: CI(412/443) = R_rs(412) / R_rs(443). Previously, the constancy of this index was confirmed for the Black Sea. In this study, using a long-term global dataset from AERONET-OC (Aerosol ROBotic NETwork-Ocean Color) stations, we show that this pattern holds across other water bodies regardless of their bio-optical characteristics. The purpose of this study is to analyse the global variability of CI(412/443) using all available AERONET-OC in situ data, to develop a method for minimising polarization-induced biases in its estimation, and to evaluate the consistency of satellite-derived CI values from major ocean colour scanners (MODIS, VIIRS, OLCI) against the corrected in situ reference.

Methods and Results. We analyse R_rs(412), R_rs(443) and their ratios obtained from AERONET-OC data. To improve the reliability of in situ data, we introduce a quantitative factor dependent on solar and viewing angles, which characterises the influence of sky polarisation on the sea surface reflectivity. A systematic underestimation of the blue colour index calculated from raw AERONET-OC data is revealed. Without accounting for viewing geometry and the degree of polarisation of skylight, the systematic error in CI can reach 0.1–0.2. Due to measurement errors in the shortwave region at low R_rs(412), raw AERONET-OC data overestimate the natural variability of the blue colour index. We estimate the maximum possible standard deviation of CI caused solely by variability in the optical properties of seawater to be 0.109. Satellite-derived CI(412/443) values at AERONET-OC station coordinates are also analysed. For all satellite sensors except Sentinel-3B OLCI, satellite CI values show high RMSD values, sometimes exceeding 0.4. In general, average CI(412/443) values from satellite data and in situ measurements show moderate agreement, although notable discrepancies exist for certain sensors and stations.

Conclusions. Our study has shown that the variability of the blue colour index across all AERONET-OC coastal stations should be significantly smaller than the variability estimated from in situ and satellite measurements. For most ocean colour scanners, variations in the blue colour index significantly exceed the expected variability. The results obtained have important practical applications for the creation and further development of regional algorithms for atmospheric correction.