Variable Masking for Climate Emulation in ACE

Abstract

In the recent AI Model Intercomparison Project (AIMIP) benchmark, the ACE2.1-ERA5 emulator was shown to underestimate the true near-surface warming trend both during the training period and the out-of-sample extrapolation period. In this work, we present M-ACE2S (Masked-ACE2S), a modified version of the stochastic ACE2S model, which uses variable masking during training and reduces the reliance of M-ACE2S on any single input channel. We show that M-ACE2S improves long-term near-surface air temperature prediction. We also present a novel data normalization strategy called "global mean removal", which, combined with masking, further improves the near-surface temperature forecasts. We show that M-ACE2S outperforms the ACE2S baseline on in-sample data for ERA5 while improving training stability and generalizability of near-surface temperature to unseen forcing when paired with global mean removal.

EC-EarthFlow: Probabilistic emulation of daily transient global climate model simulations with flow matching
Kirien Whan, Nikolaj T. Mücke, Karin van der Wiel
arXiv7 Oct 2026PreprintCC BY 4.0Full textAtmospheric science & meteorologyarXiv
Rescene: stable century-long climate emulation from a frozen weather model
Minjong Cheon
arXiv14 Oct 2026PreprintCC BY-NC-SA 4.0Full textAtmospheric science & meteorologyarXiv
Optimal scenario design for climate emulation
Christopher B. Womack, Shahine Bouabid, Andrei Sokolov +4
arXiv29 Sep 2026PreprintCC BY 4.0Full textAtmospheric science & meteorologyarXiv
METEORv1.6: Spatial climate variability and integrated impact emulation
Benjamin M. Sanderson, Marit Sandstad, Maura Dewey +2
Geoscientific Model Development30 Sep 2026Peer-reviewedCC BY 4.0Full textClimate scienceCrossref
Beyond the SCL: threshold optimization of Cloud Probability and Cloud Score+ for cloud and shadow masking in Sentinel-2 imagery using CloudSEN12
Adrián Cal, Guadalupe Tiscornia
EarthArXiv27 Oct 2026PreprintCC BY 4.0Full textGeneral earth scienceEarthArXiv

Cite

BibTeX RIS

BibTeX

@misc{yermakov2026variable,
  title = {{Variable Masking for Climate Emulation in ACE}},
  author = {Alexey Yermakov and Troy Arcomano and Jeremy McGibbon and Brian Henn and Oliver Watt-Meyer},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2610.21012},
  eprint = {2610.21012},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.21012},
  note = {CC BY 4.0},
}

RIS

TY  - UNPB
TI  - Variable Masking for Climate Emulation in ACE
AU  - Alexey Yermakov
AU  - Troy Arcomano
AU  - Jeremy McGibbon
AU  - Brian Henn
AU  - Oliver Watt-Meyer
PY  - 2026
DA  - 2026/10/14
PB  - arXiv
DO  - 10.48550/arxiv.2610.21012
UR  - https://arxiv.org/abs/2610.21012
AB  - In the recent AI Model Intercomparison Project (AIMIP) benchmark, the ACE2.1-ERA5 emulator was shown to underestimate the true near-surface warming trend both during the training period and the out-of-sample extrapolation period. In this work, we present M-ACE2S (Masked-ACE2S), a modified version of the stochastic ACE2S model, which uses variable masking during training and reduces the reliance of M-ACE2S on any single input channel. We show that M-ACE2S improves long-term near-surface air temperature prediction. We also present a novel data normalization strategy called "global mean removal", which, combined with masking, further improves the near-surface temperature forecasts. We show that M-ACE2S outperforms the ACE2S baseline on in-sample data for ERA5 while improving training stability and generalizability of near-surface temperature to unseen forcing when paired with global mean removal.
ER  -

Abstract

Related

Cite