OneScience-Group/AI-GAMFS

AI-GAMFS

Model Introduction

AI-GAMFS is a machine-learning global aerosol-meteorology forecasting system producing five-day forecasts at three-hour intervals.

Paper: Advancing operational global aerosol forecasting with machine learning
https://doi.org/10.1038/s41586-026-10234-y

Model Description

The model was proposed by teams from the Chinese Academy of Meteorological Sciences, National Meteorological Center, NASA, and collaborators. It was trained with 54 MERRA-2 aerosol and meteorological variables from 1980–2021. Vision Transformer, U-Net, and 3/6/9/12-hour relay models support global AOD, aerosol-component, and air-quality forecasts.

Use Cases

Use Case	Description
Global aerosols	Forecast AOD, optical components, and surface concentrations.
Dust and smoke	Track regional pollution transport.
Coupled weather	Jointly forecast aerosols and meteorology.
ModelScope/OneCode execution	Validate data, training, inference, aerosol metrics, and visualization.
Multi-GPU training	Start multi-process training through torchrun.

Usage Instructions

Use a GPU or DCU when available; CPU supports the default smoke configuration.

hf download OneScience-Group/AI-GAMFS --local-dir ./AI-GAMFS
cd AI-GAMFS

Environment Dependencies

Hardware Requirements

• A GPU or DCU is recommended.
• A CPU can be used for connectivity validation with the default small-sample configuration.
• DCU users should install DTK 25.04.2 or a compatible OneScience-recommended version first.

DCU Environment

# Activate DTK and Conda first
conda create -n onescience311 python=3.11 -y
conda activate onescience311
pip install onescience[earth-dcu] -i http://mirrors.onescience.ai:3141/pypi/simple/ --trusted-host mirrors.onescience.ai

GPU Environment

# Activate Conda first
conda create -n onescience311 python=3.11 -y libstdcxx-ng=12 libgcc-ng=12 gcc_linux-64=12 gxx_linux-64=12
conda activate onescience311
pip install onescience[earth-gpu] -i http://mirrors.onescience.ai:3141/pypi/simple/ --trusted-host mirrors.onescience.ai

python scripts/fake_data.py
python scripts/train.py
torchrun --standalone --nproc_per_node=2 scripts/train.py
python scripts/inference.py
python scripts/result.py

Training optimizes four relay models. Inference produces 40 three-hourly forecasts and evaluation reports finite AOD RMSE and correlation.

Trained Weights

No weights are bundled under weight/. The paper does not provide a directly loadable official pretrained-weight link.

Citation and License

This repository is an independent engineering reproduction of the public AI-GAMFS specifications.

The original paper is licensed under CC BY-NC-ND 4.0; official code, model weights, and related data retain their respective terms.