Towards fully differentiable neural ocean model with Veros
This work addresses the problem of enabling end-to-end learning and parameter tuning in ocean modeling for researchers and scientists, representing an incremental advancement by applying differentiable programming to an existing model.
The researchers tackled the challenge of making ocean models differentiable by extending the VEROS model to be compatible with JAX, enabling automatic differentiation through its dynamical core. They demonstrated applications such as correcting initial ocean states and calibrating physical parameters, showing how this facilitates gradient-based optimization and parameter tuning in ocean modeling.
We present a differentiable extension of the VEROS ocean model, enabling automatic differentiation through its dynamical core. We describe the key modifications required to make the model fully compatible with JAX autodifferentiation framework and evaluate the numerical consistency of the resulting implementation. Two illustrative applications are then demonstrated: (i) the correction of an initial ocean state through gradient-based optimization, and (ii) the calibration of unknown physical parameters directly from model observations. These examples highlight how differentiable programming can facilitate end-to-end learning and parameter tuning in ocean modeling. Our implementation is available online.