Hybrid physics-ML & parameterization
Explainable Offline-Online Training of Neural Networks for Parameterizations: A 1D Gravity Wave-QBO Testbed in the Small-data Regime
In one sentence. Online re-training of two layers of a gravity-wave CNN via ensemble Kalman inversion yields realistic QBOs in a 1D model when offline small-data training fails.
Abstract
There are different strategies for training neural networks (NNs) as subgrid-scale parameterizations. Here, we use a 1D model of the quasi-biennial oscillation (QBO) and gravity wave (GW) parameterizations as testbeds. A 12-layer convolutional NN that predicts GW forcings for given wind profiles, when trained offline in a big-data regime (100-years), produces realistic QBOs once coupled to the 1D model. In contrast, offline training of this NN in a small-data regime (18-months) yields unrealistic QBOs. However, online re-training of just two layers of this NN using ensemble Kalman inversion and only time-averaged QBO statistics leads to parameterizations that yield realistic QBOs. Fourier analysis of these three NNs' kernels suggests why/how re-training works and reveals that these NNs primarily learn low-pass, high-pass, and a combination of band-pass filters, consistent with the importance of both local and non-local dynamics in GW propagation/dissipation. These findings/strategies apply to data-driven parameterizations of other climate processes generally.
Labels
- Topic
- Hybrid physics-ML & parameterization, Explainability, uncertainty & physics
- Type
- New model or method
- Method
- CNN / U-Net, Physics-informed / physics-constrained
- Time scale
- Climate projections / centennial
- Domain
- Local / station
- Data
- Other
- Status
- Research
- Identifiers
- arXiv:2309.09024 DOI:10.1029/2023gl106324 OpenAlex:W4391276972
- Added
- 2026-10-22
Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.