AERIS

Sam Foreman

AERIS: Argonne Earth Systems Model for Reliable and Skillful Predictions

High-Level Overview of AERIS

Figure 2: Rollout of AERIS model, specific humidity at 700m.

Table 1: Overview of AERIS model and training setup

Property	Description
Domain	Global
Resolution	0.25° & 1.4°
Training Data	ERA5 (1979–2018)
Model Architecture	Swin Transformer
Speedup¹	O(10k–100k)

Model Overview

Dataset: ECMWF Reanalysis v5 (ERA5)
Variables: Surface and pressure levels
Usage: Medium-range weather forecasting
Partition:
- Train: 1979–2018
- Val: 2019
- Test: 2020
Data Size: 100GB at 5.6° to 31TB at 0.25°

Table 2: Variables used in AERIS training and prediction

Variable	Description
`t2m`	2m Temperature
`X` `u`(`v`)	$u$ ( $v$ ) wind component @ Xm
`q`	Specific Humidity
`z`	Geopotential
`msl`	Mean Sea Level Pressure
`sst`	Sea Surface Temperature
`lsm`	Land-sea mask

Windowed Self-Attention

Benefits for weather modeling:
- Shifted windows capture both local patterns and long-range context
- Constant scale, windowed self-attention provides high-resolution forecasts
- Designed (currently) for fixed, 2D grids
Inspiration from SOTA LLMs:
- RMSNorm, SwiGLU, 2D RoPE

Model Architecture: Details

Sequence-Window-Pipeline Parallelism `SWiPe`

SWiPe is a novel parallelism strategy for Swin-based Transformers
Hybrid 3D Parallelism strategy, combining:
- Sequence parallelism (SP)
- Window parallelism (WP)
- Pipeline parallelism (PP)

Figure 6: `SWiPe` Communication Patterns

🌌 Aurora

Table 3: Aurora Specs


Racks	166
Nodes	10,624
CPUs	21,248
GPUs	63,744
NICs	84,992
HBM	8 PB
DDR5c	10 PB

🌎 AERIS: Scaling Results

10 EFLOPs (sustained) @ 120,960 GPUs
See (Hatanpää et al. (2025)) for additional details
arXiv:2509.13523

References

Hatanpää, Väinö, Eugene Ku, Jason Stock, Murali Emani, Sam Foreman, Chunyong Jung, Sandeep Madireddy, et al. 2025. “AERIS: Argonne Earth Systems Model for Reliable and Skillful Predictions.” https://arxiv.org/abs/2509.13523.