Global retrieval of phytoplankton phenology and primary production using DINEOF reconstructed satellite chlorophyll a product

Time-series interpretation of ocean color data is often necessarily restricted by large data gaps due to cloud cover, sea ice and high solar zenith angle. Among several methods of filling in missing data, Data Interpolation Empirical Orthogonal Function (DINEOF; Becker & Rixen, 2003) is an effective approach for incomplete oceanographic datasets on physical variables (e.g. sea surface temperature and sea surface salinity). However, to our knowledge, evaluations have not yet been conducted enough regarding the applicability of the DINEOF to ocean color data on biogeochemical variables (e.g. chlorophyll a concentration). In this study, we perform an error analysis using complete and global gridded Chl a concentration data from the ECCO-Darwin model in order to examine to what extent the DINEOF can be applied to ocean color data in terms of the number of observations in a given image. Within an acceptable threshold determined from the above-mentioned analysis, we apply the DINEOF on datasets on daily global Chl a concentration obtained from Aqua/MODIS for 2003-2020. We illustrate that spatiotemporal changes in phytoplankton biomass and the associated primary production can be detected using the reconstructed data, which is consistent with finding in the literature.