Scientists break ‘decades of gridlock’ in climate modeling

Finally, there is hope for better and more accurate weather forecasting!

Maybe soon, the latest climate models will confirm that the Global Warming/Climate Change was a hoax!

"In brief

• Gravity waves are a source of uncertainty in climate models because they are too small and short-lived to appear in models designed to cover the whole planet.
• A new Stanford-led study shows how machine learning algorithms that predict the effects of gravity waves can be incorporated into global climate models.
• The approach shows a path toward better modeling of other small-scale systems, like clouds, and could improve understanding of future weather patterns.

...

Climate models don’t fully capture gravity waves because they are often based on a grid of 100-by-100-kilometer square columns. In each column, physics equations describe the movement of air and water. Many gravity waves are too small to register at this resolution, like a ripple in a puddle that a low-resolution photo doesn’t capture. Other gravity waves ripple out over distances long enough to cross 10 or more squares in the grid. But, due to computational constraints, climate models do not capture horizontal gravity wave movement.  ..."

From the 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  ANN, a  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 

 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."

Scientists break ‘decades of gridlock’ in climate modeling | Stanford Report "In global climate models, researchers have harnessed AI to accurately model atmospheric gravity waves, ripples of air that affect the polar vortex, winter weather, and climate patterns."

Offline Performance of a Nonlocal Deep Learning Parameterization for Climate Model Representation of Atmospheric Gravity Waves (open access)

Fig. 1 (left) Temperature perturbations (in K) associated with gravity waves (GWs) over the Drake Passage and the Southern Ocean on 18 July 2015 06 UTC, as resolved in ERA5,

(middle) the momentum flux  (units mPa) associated with the excited GWs, and (right) the momentum flux predicted using an Attention UNet convolutional neural network trained on 3 years of ERA5 data.