allenai/ACE2S-SHiELD-plus · Hugging Face

Log In<br>Sign Up

ACE2S-SHiELD+

Ai2 Climate Emulator (ACE) is a family of models designed to simulate<br>atmospheric variability from the time scale of days to centuries.

Disclaimer: ACE models are research tools and should not be used for<br>operational climate predictions.

ACE2S-SHiELD+ is an emulator of GFDL's physics-based SHiELD model, trained on a<br>combination of AMIP, equilibrium-climate, and ramped-SST-random-CO2<br>data. It has comparable skill to ACE2-SHiELD in AMIP inference and ACE2-SOM in<br>slab-ocean-coupled equilibrium-climate inference, but also can accurately<br>emulate scenarios with independent perturbations to the SST or CO2,<br>like AMIP +4 K or slab-ocean-coupled abrupt 4xCO2. It is a<br>stochastic model, which facilitates running large ensembles of simulations from<br>the same initial conditions, has an improved representation of the spherical<br>power spectrum of its predicted variables, and also includes a new constraint<br>to conserve global atmospheric total energy.

Quick links:

📃 Paper

💻 Code

💬 Docs

📂 All Models

Briefly, the strengths of ACE2S-SHiELD+ are:

It can accurately perform historical AMIP simulations, slab-ocean-coupled<br>simulations with constant or steadily increasing CO2 between 1x<br>and 4x the concentration of the present day.

Unlike prior models, it can additionally accurately separate the independent<br>effects of SST and CO2 on climate, including in AMIP +4 K and<br>abrupt 4xCO2 inference, notably emulating the correct radiative<br>sensitivity to changes in CO2 including the implicit response of<br>clouds.

In addition to conserving dry air mass and water like prior models, it also<br>is constrained to conserve global atmospheric total energy within the same<br>average residual of SHiELD.

Some known weaknesses are:

It is trained to emulate a physics-based model and therefore inherits the<br>biases relative to observations thereof.

As in the case of ACE2-SOM and SHiELD, in slab-ocean mode, sea-ice coverage,<br>ocean heat transport, and ocean mixed layer depth are prescribed based on a<br>present-day climatologies and therefore do not respond to changes in<br>CO2. Prescribed sea-ice means projections with ACE2S-SHiELD+ are<br>missing an important feedback mechanism known to amplify warming in the polar<br>regions. Due to the narrow spread in sea ice coverage in samples seen during<br>training, generalization ability to unseen sea ice coverage is limited.

Similar to ACE2-SOM, abrupt regime shifts in stratospheric temperature and<br>moisture can occur in inference runs, which sometimes can affect predictions<br>of other variables.

Inference speed:

Note that due to architectural and hyperparameter changes, inference speed with<br>ACE2S-SHiELD+ is roughly half that of ACE2-SHiELD or ACE2-SOM, but it is still<br>68x faster than SHiELD when each are run on typical hardware.

License

This model is licensed under Apache 2.0. It is intended for research and<br>educational use in accordance with Ai2’s<br>Responsible Use Guidelines.

Downloads last month 8

Inference Providers NEW<br>This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Paper for allenai/ACE2S-SHiELD-plus<br>Paper • 2606.07928 • Published 10 days ago • 1