Assessing chlorophyll-a retrieval of Sentinel-3 OLCI, VIIRS, MODIS and OC-CCI in Monterey Bay (California, USA)

ABSTRACT Advances in high-resolution satellite sensors and merged multi-sensor ocean colour products have improved the detection of submesoscale features and enhanced the continuity of ocean surface measurements. This has enabled more detailed observations of chlorophyll-a (chl-a) growth rates, biomass and biogeochemical fluxes. Chl-a, a photosynthetic pigment in marine autotrophs, is used as a proxy for phytoplankton biomass and primary productivity. The accurate estimation of surface chl-a via remote sensing is key to improving coastal ocean process modelling. We compared chl-a products from sensors of multiple spatial resolutions, including Sentinel-3 Ocean and Land Colour Imager (OLCI), the Moderate Resolution Imaging Spectroradiometer (Aqua) (MODIS Aqua), the Visible Infrared Imaging Radiometer Suite (VIIRS) on Suomi NPP and the Ocean Colour Climate Change Initiative (OC-CCI) V6 product to two in situ datasets (M1 and C1) in Monterey Bay, California, USA. Sentinel-3 consistently performed well under the Inverse Modelling Technique - Neural Network chl-a algorithm, performing slightly better at station M1 R M 1 2 = 0 . 85 , R C 1 2 = 0 . 80. The Ocean Colour for MERIS algorithm performed better at offshore station M1 R M 1 2 = 0 . 87 and yielded no matchups at C1. MODIS Aqua and VIIRS performed poorly R M 1 , M O D I S 2 = 0 . 32 , R M 1 , V I I R S 2 = 0 . 02 , R C 1 , M O D I S 2 = 0 . 08 , R C 1 , V I I R S 2 = 0 . 25, likely inhibited by calibration to global as opposed to regionally adapted datasets. OC-CCI, though blended for greater continuity, produced fewer matchups at a higher accuracy R M 1 , O C - C C I 2 = 0 . 66 , R C 1 , O C - C C I 2 = 0 . 63 than MODIS Aqua or VIIRS alone, but still less so than Sentinel-3 OLCI. All sensors either over- or under-estimated chl-a at all concentrations, apart from OC-CCI, likely due to the complex coastal optical properties of the area. Our results highlight the need for regional algorithm development and show the potential effectiveness of Sentinel-3 and OC-CCI products for future submesoscale process studies.