Physics-Informed Deep Learning for Subsurface Current Estimation in the Florida Straits: Toward Operational Resource Assessment for Ocean Current Turbines
Authors/Creators
Description
Ocean current turbines (OCTs) deployed in persistent, fast-flowing western boundary currents represent a promising renewable energy resource with strong potential to support coastal electrification along the southeastern United States. OCT deployment requires accurate characterization of three-dimensional currents within the upper 100–150 m, where turbines are expected to operate. Conventional methods rely on sparse ADCP observations or computationally intensive ocean circulation models, limiting rapid and spatially resolved operational assessment. Deep learning can reconstruct subsurface ocean states from surface observations, but data-only models may produce physically inconsistent currents away from the surface. This work develops a physics-informed framework that estimates three-dimensional subsurface velocities in the Florida Straits from 35 gridded surface and hydrographic input channels. The objective is to improve depth-resolved Florida Current and transport reconstruction while retaining the flexibility of data-driven learning for ocean-current-turbine resource assessment.