Abstract
Gravity waves (GWs) make crucial contributions to the middle atmospheric circulation. Yet, their climate model representation remains inaccurate, leading to key circulation biases. This study introduces a set of three neural networks (NNs) that learn to predict GW fluxes (GWFs) from multiple years of high-resolution ERA5 reanalysis. The three NNs: a (Formula presented.) ANN, a (Formula presented.) ANN-CNN, and an Attention UNet embed different levels of horizontal nonlocality in their architecture and are capable of representing nonlocal GW effects that are missing from current operational GW parameterizations. The NNs are evaluated offline on both time-averaged statistics and time-evolving flux variability. All NNs, especially the Attention UNet, accurately recreate the global GWF distribution in both the troposphere and the stratosphere. Moreover, the Attention UNet most skillfully predicts the transient evolution of GWFs over prominent orographic and nonorographic hotspots, with the (Formula presented.) model being a close second. Since even ERA5 does not resolve a substantial portion of GWFs, this deficiency is compensated by subsequently applying transfer learning on the ERA5-trained ML models for GWFs from a 1.4 km global climate model. It is found that the re-trained models both (a) preserve their learning from ERA5, and (b) learn to appropriately scale the predicted fluxes to account for ERA5's limited resolution. Our results highlight the importance of embedding nonlocal information for a more accurate GWF prediction and establish strategies to complement abundant reanalysis data with limited high-resolution data to develop machine learning-driven parameterizations for missing mesoscale processes in climate models.
| Original language | English |
|---|---|
| Article number | e2025MS004977 |
| Journal | Journal of Advances in Modeling Earth Systems |
| Volume | 17 |
| Issue number | 10 |
| DOIs | |
| State | Published - Oct 2025 |
Funding
Aditi Sheshadri and Aman Gupta are supported by Schmidt Sciences, LLC, a philanthropic initiative founded by Eric and Wendy Schmidt, as part of the Virtual Earth System Research Institute (VESRI). Aditi Sheshadri also acknowledges support from the National Science Foundation through Grant OAC-2004492. The work was also supported by NASA's Office of Chief Science Data Officer and the Earth Science Division's Earth Science Scientific Computing, Earth Science Data Systems Program, and the Earth Science Modeling and Analysis Program. The 1.4 km IFS runs were performed by Nils Wedi and Inna Polichtchouk (ECMWF) on the Summit Supercomputer at the Oak Ridge National Laboratory, using resources of the Oak Ridge Leadership Computing Facility (OLCF), which is a DOE Office of Science User Facility supported under Contract DE-AC05-00OR22725. We thank M. Joan Alexander, Edwin P. Gerber, Pedram Hassanzadeh, Laura Mansfield, Hamid Pahlavan, Y. Qiang Sun, Brian Green, Robert King, Manil Maskey, and Rahul Ramachandran for insightful discussions. We also thank two anonymous reviewers for helpful comments on the transfer learning experiments. Aditi Sheshadri and Aman Gupta are supported by Schmidt Sciences, LLC, a philanthropic initiative founded by Eric and Wendy Schmidt, as part of the Virtual Earth System Research Institute (VESRI). Aditi Sheshadri also acknowledges support from the National Science Foundation through Grant OAC‐2004492. The work was also supported by NASA's Office of Chief Science Data Officer and the Earth Science Division's Earth Science Scientific Computing, Earth Science Data Systems Program, and the Earth Science Modeling and Analysis Program. The 1.4 km IFS runs were performed by Nils Wedi and Inna Polichtchouk (ECMWF) on the Summit Supercomputer at the Oak Ridge National Laboratory, using resources of the Oak Ridge Leadership Computing Facility (OLCF), which is a DOE Office of Science User Facility supported under Contract DE‐AC05‐00OR22725. We thank M. Joan Alexander, Edwin P. Gerber, Pedram Hassanzadeh, Laura Mansfield, Hamid Pahlavan, Y. Qiang Sun, Brian Green, Robert King, Manil Maskey, and Rahul Ramachandran for insightful discussions. We also thank two anonymous reviewers for helpful comments on the transfer learning experiments.
Keywords
- atmospheric dynamics and variability
- climate model parameterizations
- gravity waves
- machine learning
- nonlocal physical parameterizations
- stratospheric circulation