What question did this study set out to answer?

To categorize and evaluate methods for embedding physical laws into machine learning models to improve weather and climate simulation, downscaling, and forecasting.

February 15, 2021Philosophical Transactions of the Royal Society A Mathematical Physical and Engineering Sciences

Physics-informed machine learning: case studies for weather and climate modelling

View Full Paper

Ask AI

Bookmark

To categorize and evaluate methods for embedding physical laws into machine learning models to improve weather and climate simulation, downscaling, and forecasting.
Surveyed methodological frameworks designed to incorporate domain knowledge and governing physical principles into machine learning architectures.
Evaluated 10 multi-application case studies focused on emulating, downscaling, and predicting spatio-temporal weather and climate phenomena.
Synthesized computational, scientific, and diagnostic challenges confronting the deployment of physics-informed models.
Physics-informed machine learning models achieved superior physical consistency, improved out-of-distribution generalization, and higher data efficiency compared to standard architectures.
Incorporating physical constraints reduced model training duration while maintaining predictive accuracy across nonlinear weather and climate tasks.
Analysis identified key bottlenecks in scientific diagnostics, computational resource scaling, and systematic validation required for operational reliability.

Discussion

No takes yet. Share an insight, caveat, or question.

Implication

Narrative review demonstrates enhanced physical consistency and generalization across weather and climate modelling, highlighting the value of physics-informed machine learning.

Cite This Study

Kashinath et al. (2021) studied this question.

synapsesocial.com/papers/69d88f188c03fbaff8bef642 https://doi.org/10.1098/rsta.2020.0093

View Full Paper

Ask AI

Bookmark

Implication

Narrative review demonstrates enhanced physical consistency and generalization across weather and climate modelling, highlighting the value of physics-informed machine learning.

Physics-informed machine learning: case studies for weather and climate modelling

Key Points

Discussion

Implication

Cite This Study

Physics-informed machine learning: case studies for weather and climate modelling

Key Points

Discussion

Implication

Cite This Study

Also Consider

Also Consider