Ocean colour as a tracer for surface current reconstruction: An approach using a physics informed neural network

Ocean color has been used to monitor phytoplankton biomass and ecosystem processes; thanks to its sharp gradients it can also provide a reliable source of data as a tracer for surface dynamics. The growing availability of high resolution products (e.g Sentinel-2, Sentinel-3), offers the potential of capturing processes at different spatial scales, from large scale oceanic processes to eddies and filaments, resolving meso- to submesoscale features, complementing the information extracted from other established tracers as sea surface temperature. While traditional methods for estimating surface currents from remote sensing rely on altimetry and the geostrophic approximation, these approaches face limitations in resolving finer scales.

In this study, we explore the capabilities of a neural network to reconstruct ocean surface currents based on the advection visible in ocean colour tracer fields and auxiliary variables such as altimetry products. We implemented a convolutional encoder-decoder architecture trained with a cost function that minimizes the difference between advected tracer fields and their observed distributions, with additional constraints to fit physical boundaries (coastal boundaries) and flow properties (divergence). This approach aims to overcome limitations such as sparse along-track sampling and revisit times, often observed in traditional altimetry products, by exploiting the finer spatial and temporal resolution of ocean colour data.

In this first stage we used model fields from the Irish-Iberian Biscay (IBI-MFC) reanalysis and non-assimilative biogeochemical hindcast, testing chlorophyll, temperature and salinity as input fields. Preliminary results indicate that chlorophyll fields capture variability associated with mesoscale, while salinity provides more stable reconstructions. The next stage involves using multiple tracers as input fields to exploit the merged ocean colour products and obtain finer scale outputs for ocean surface processes.