We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
FaIRGP: A Bayesian Energy Balance Model for Surface Temperatures Emulation.
- Authors
Bouabid, Shahine; Sejdinovic, Dino; Watson‐Parris, Duncan
- Abstract
Emulators, or reduced complexity climate models, are surrogate Earth system models (ESMs) that produce projections of key climate quantities with minimal computational resources. Using time‐series modeling or more advanced machine learning techniques, data‐driven emulators have emerged as a promising avenue of research, producing spatially resolved climate responses that are visually indistinguishable from state‐of‐the‐art ESMs. Yet, their lack of physical interpretability limits their wider adoption. In this work, we introduce FaIRGP, a data‐driven emulator that satisfies the physical temperature response equations of an energy balance model. The result is an emulator that (a) enjoys the flexibility of statistical machine learning models and can learn from data, and (b) has a robust physical grounding with interpretable parameters that can be used to make inference about the climate system. Further, our Bayesian approach allows a principled and mathematically tractable uncertainty quantification. Our model demonstrates skillful emulation of global mean surface temperature and spatial surface temperatures across realistic future scenarios. Its ability to learn from data allows it to outperform EBMs, while its robust physical foundation safeguards against the pitfalls of purely data‐driven models. We also illustrate how FaIRGP can be used to obtain estimates of top‐of‐atmosphere radiative forcing and discuss the benefits of its mathematical tractability for applications such as detection and attribution or precipitation emulation. We hope that this work will contribute to widening the adoption of data‐driven methods in climate emulation. Plain Language Summary: Emulators are simplified climate models that can be used to rapidly explore climate scenarios—they can run in less than a minute on an average computer. They are key tools used by the Intergovernmental Panel on Climate Change to explore the diversity of possible future climates. Data‐driven emulators use advanced machine learning techniques to produce climate predictions that look very similar to the predictions of complex climate models. However, they are not easy to interpret, and therefore to trust in practice. In this work, we introduce FaIRGP, a data‐driven emulator based on physics. The emulator is flexible and can learn from data to improve its predictions, but is also grounded on physical energy balance relationships, which makes it robust and interpretable. The model performs well in predicting future global and local temperatures under realistic future scenarios, outperforming purely physics‐driven or purely data‐driven models. Further, the probabilistic nature of our model allows for mathematically tractable uncertainty quantification. By gaining trust in such a data‐driven yet physically grounded model, we hope the climate science community can benefit more widely from their potential. Key Points: We introduce FaIRGP, a Bayesian machine learning emulator for global and local mean surface temperatures that builds upon a physically based simple climate modelThe model improves upon both purely physically‐driven and purely data‐driven baseline emulators on several metrics across realistic future scenariosThe model is fully mathematically tractable, which makes it a convenient and easy‐to‐use probabilistic tool for the emulation of surface temperatures, but also for downstream applications such as detection and attribution or precipitation emulation
- Subjects
INTERGOVERNMENTAL Panel on Climate Change; SURFACE temperature; MACHINE learning; STATISTICAL learning; CLIMATOLOGY; ATTRIBUTION (Social psychology); ATMOSPHERIC models; RADIATIVE forcing
- Publication
Journal of Advances in Modeling Earth Systems, 2024, Vol 16, Issue 6, p1
- ISSN
1942-2466
- Publication type
Article
- DOI
10.1029/2023MS003926