Reconstructing mesoscale surface circulation in the Bay of Biscay using an advection-based neural network

Ocean mesoscale dynamics play a major role in phytoplankton productivity and other physically driven ecological processes in the Bay of Biscay, a region characterized by complex bathymetry and a large-scale poleward current that promotes the formation of slope water oceanic eddies (SWODDIES). These features strongly influence the horizontal transport and distribution of tracers, contributing to a highly variable biophysical environment. However, the role of submesoscale to mesoscale circulation in shaping chlorophyll distribution remains difficult to quantify, partly due to the limited characterization of surface currents at these spatial scales.
High-resolution ocean color and sea surface temperature datasets capture fine-scale tracer variability and provide an opportunity to infer surface circulation through an inverse approach. In this study, we explore the capability of a neural network to reconstruct ocean surface currents and mesoscale features in the Bay of Biscay based on the advection visible in tracer fields, complemented by auxiliary variables such as altimetry products.
We implemented a convolutional encoder–decoder architecture trained with an advection-based cost function that minimizes the difference between advected tracer fields and their observed distributions, while incorporating constraints on physical boundaries and flow properties such as divergence. The method is evaluated using daily fields from the Irish–Iberian Biscay (IBI-MFC) reanalysis and a non-assimilative biogeochemical hindcast for the period 1993–2019.
The network is then used to reconstruct velocity fields for the year 2020, showing consistent spatial patterns along selected transects, with an RMSE of 0.039 m/s compared to a mean velocity of 0.059 m/s. This approach highlights the potential of tracer observations to improve our understanding of mesoscale surface circulation and transport processes in the Bay of Biscay.