Surrogate Neural Networks to Estimate Parametric Sensitivity of Ocean Models
This work addresses the challenge of unmeasurable parameters affecting ocean model outputs, which is crucial for understanding climate impacts like greenhouse gases and hurricanes, but it is incremental as it applies existing surrogate methods to a specific domain.
The authors tackled the problem of estimating parametric sensitivity in ocean models by training surrogate neural networks on perturbed parameter ensemble data, achieving accurate predictions of one-step forward dynamics.
Modeling is crucial to understanding the effect of greenhouse gases, warming, and ice sheet melting on the ocean. At the same time, ocean processes affect phenomena such as hurricanes and droughts. Parameters in the models that cannot be physically measured have a significant effect on the model output. For an idealized ocean model, we generated perturbed parameter ensemble data and trained surrogate neural network models. The neural surrogates accurately predicted the one-step forward dynamics, of which we then computed the parametric sensitivity.