|
High-quality satellite-based ocean color products derived from Sentinel-2/MSI and Sentinel-3/OLCI provide valuable support and insights in the management and monitoring of coastal ecosystems. The primary ocean color variable is the spectral Remote Sensing Reflectance (RRS), obtained after applying atmospheric correction (AC) on satellite products. AC algorithms, such as C2RCC and ACOLITE/DSF are all well capable of generating RRS products over coastal waters. The question of which approach to choose is important and not obvious, especially considering different water (e.g. turbid, clear or CDOM rich waters) and atmospheric conditions (e.g. sun glint, low sun angles) which can occur in coastal waters. To improve the operational ability to achieve high quality RRS spectra for a maximum number of pixels and yet retain the ability to deal with both unusual water conditions and challenging atmospheric conditions, we present the merged use of two algorithms: C2RCC and ACOLITE/DSF. Combining the two approaches yet required a comprehensive, region independent and pixel-based automatic switching scheme, along with a technique for achieving a seamless transition between the two algorithms. We here used the green-NIR ratio, which offers a clear indication of the saturation of the C2RCC outputs for the most reflective band (i.e., the RRS560), at a level where ACOLITE/DSF typically performs accurately, combined with a weighted transition between the two methods. The approach was applied to both Sentinel-2/MSI and Sentinel-3/OLCI products and validated using autonomous WATERHYPERNET stations located in Oostende (RT1, Belgium) and Venice (AAOT, Italy), showing an improved quality of the RRS products compared to using the ACs independently. The best results are obtained for the merged approach in the bands 443nm to 709nm for both Sentinel-2/MSI (<21% MAPE with a 0.004 RMSD and slopes between 0.93 and 0.98) and Sentinel-3/OLCI (<23% MAPE with a 0.003 RMSD and slopes between 0.91 and 0.98) which have generally the highest reflectance range, and which are generally of interest to retrieve turbidity in low to moderately turbid waters.
|
