Physics-informed neural networks bridge the gap between satellite ocean temperatures and subsurface coral conditions, enabling more accurate thermal stress assessments for bleaching monitoring.

Satellite SST measurements only capture surface temperatures but underestimate subsurface thermal stress, with depths beyond 20 metres potentially 1-3°C cooler than the surface

A PINN model successfully fuses NOAA Coral Reef Watch SST data with sparse in-situ temperature loggers using the vertical heat equation, achieving 0.25-1.38°C RMSE accuracy at unseen depths

The model learns effective thermal diffusivity and light attenuation parameters while enforcing satellite SST as a hard boundary condition

Tested across four Great Barrier Reef sites with 30 holdout experiments, the PINN maintains strong performance (0.27-0.32°C RMSE) even with extreme data sparsity of only three training depths