Maps of time-mean precipitation pattern error for 40-day simulations with three configurations of a global atmospheric model with a coarse 200-km grid.
Maps of time-mean precipitation pattern error for 40-day simulations with three configurations of a global atmospheric model with a coarse 200-km grid. Error is measured with respect to a 40-day ‘truth’ simulation using a similar model with a 3 km global grid. RMSE is a global average of the pattern error. Errors of a ‘baseline’ model (panel a) are reduced by 30% by adding a correction to that model’s physical parameterizations machine-learned using (b) a random forest, or (c) a neural net, trained on output from the 3 km simulation. Panel (d) shows that both machine learning corrections also improve the partitioning of precipitation between land and ocean. Credit: Bretherton et al. [2022], Figure 11
Editors’ Highlights are summaries of recent papers by AGU’s journal editors.
Source: Journal of Advances in Modeling Earth Systems

A new generation of global atmospheric models with 1 to 5 kilometer grid spacing can accurately simulate extreme weather, such as the local heavy rain from thunderstorm systems and flow across complex mountain ranges. They could also help us better plan for local-scale impacts of future climate change, but they are far too computation-heavy to use for simulations of decades or centuries. Bretherton et al. [2022] present a machine learning approach to correcting the coarser-grid climate models that we can afford to run using outputs from short reference simulations with fine-grid climate models. The correction makes the coarse-grid model more closely track weather forecasts and time-mean rainfall patterns from the reference simulation.

Citation: Bretherton, C. S., Henn, B., Kwa, A., Brenowitz, N. D., Watt-Meyer, O., McGibbon, J., et al. (2022). Correcting coarse-grid weather and climate models by machine learning from global storm-resolving simulations. Journal of Advances in Modeling Earth Systems, 14, e2021MS002794. https://doi.org/10.1029/2021MS002794

―Jiwen Fan, Editor, JAMES

Text © 2022. The authors. CC BY-NC-ND 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.