9  Physics-based models and calibration 🚧

Learning objectives

After reading this chapter, you should be able to:

• Navigate trade-offs between model complexity, interpretability, and computational cost
• Characterize and communicate within- and between-model uncertainty
• Use surrogate models to approximate complex model output
• Apply model calibration techniques for climate applications

9.1 Essential concepts

1. Physics-based vs data-driven modeling spectrum
2. Model chaining and uncertainty propagation
3. Calibration methods and parameter estimation
4. Surrogate modeling for computational efficiency
5. Model structure uncertainty quantification

Further reading

For physics-based modeling in climate applications:

• Rackauckas et al. (2020): Scientific machine learning for differential equations

Rackauckas, Christopher, Yingbo Ma, Julius Martensen, Collin Warner, Kirill Zubov, Rohit Supekar, Dominic Skinner, Ali Ramadhan, and Alan Edelman. 2020. “Universal Differential Equations for Scientific Machine Learning.” 2020. https://doi.org/10.48550/ARXIV.2001.04385.