A machine learning approach to produce a continuous solar-induced chlorophyll fluorescence dataset for understanding Ocean productivity

Phytoplankton primary production is a crucial component of Arctic Ocean (AO) biogeochemistry, playing a pivotal role in the carbon cycling by supporting higher trophic levels and removing atmospheric carbon dioxide. The advent of satellite observations measuring chlorophyll a concentration (Chl_ a) has yielded unprecedented insights into the distribution of AO phytoplankton, enhancing our ability to assess oceanic productivity. However, the optical properties of AO waters differ significantly from those of lower‐latitude waters, and standard Chl_a algorithms perform poorly in the AO. In particular, Chl_a retrievals are challenged by interferences from other marine constituents including higher pigment packaging and higher proportion of light absorption by colored dissolved organic matter. To derive phytoplankton-originating signature as well as mitigate those effects, solar-induced chlorophyll fluorescence (SIF) emerges as a valuable tool for acquiring physiological insights into the direct photosynthetic processes in the AO. In this study, we leverage satellite-based SIF measurements to assess their correlation with a set of predictive factors influencing phytoplankton photosynthesis. We extend the temporal coverage of AO SIF data to cover the period 2004 - 2020. This novel dataset offers a pathway to monitor the physiological interactions of phytoplankton with changes in climate, promising to significantly improve our understanding of the Arctic water’s productivity. The application of this data is expected to provide insights into how phytoplankton respond to shifts in environmental changes, contributing to a more nuanced understanding of their role in High-Latitude Northern Oceans ecosystems.