Skip to content

OAE PINN Emulator

A physics-informed neural network emulator for 2-D ocean tracer transport used in Ocean Alkalinity Enhancement (OAE) studies.

The project explores machine learning based emulators for tracer evolution using Oceananigans.jl simulations as training while also penalizing advection–diffusion PDE residuals, aiming to stay physically consistent and generalize across forcing.

View code Run demo

Example prediction (placeholder)

True vs predicted tracer

Model overview

PINN architecture

Tech stack

  • Oceananigans.jl simulations → tracer + velocity fields
  • NetCDF outputs → processed with xarray
  • Training in PyTorch (single-step + multi-step variants; physics loss optional)