OneScience-Group/CLM5-Emulator

CLM5-Emulator

Model Introduction

CLM5-Emulator uses machine learning to emulate global biophysical responses of Community Land Model version 5. Six parameters drive predictions of GPP and LHF EOF components and bounded parameter estimation.

Paper: A machine learning approach to emulation and biophysical parameter estimation with the Community Land Model, version 5
https://doi.org/10.5194/ascmo-6-223-2020

Model Description

The model was proposed by researchers at NCAR and CERFACS. It was trained with GSWP3-driven CLM5 parameter perturbation ensembles and FLUXNET-MTE observational targets. Two independent feed-forward networks emulate GPP and LHF spatial components for surrogate modeling and biophysical parameter estimation.

Use Cases

Use Case	Description
CLM5 emulation	Predict GPP and LHF components from six parameters.
Parameter estimation	Search the bounded normalized parameter space.
Spatial reconstruction	Reconstruct global responses from EOF components.
ModelScope/OneCode execution	Validate structured data, training, inference, metrics, and visualization.
Multi-GPU training	Start multi-process training through torchrun.

Usage Instructions

hf download OneScience-Group/CLM5-Emulator --local-dir ./CLM5-Emulator
cd CLM5-Emulator

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 --nproc_per_node=2 --nnodes=1 --master_addr="localhost" --master_port=29500 scripts/train.py
python scripts/inference.py
python scripts/result.py

Synthetic data retain [B,6] inputs, independent GPP/LHF targets, three EOF modes, and the logical 4-by-5-degree grid. Both target networks participate in backpropagation, and single-process and two-process DDP training have been verified. Inference restores the checkpoint, produces components with shape [8,2,3] and fields with shape [8,2,46,72], and verifies finite outputs. Training results are saved to result/checkpoints/clm5_emulator.pt; inference and evaluation results are saved under result/output/ and result/evaluation/.

Official OneScience Information

Platform	OneScience Main Repository	Skills Repository
Gitee	https://gitee.com/onescience-ai/onescience	https://gitee.com/onescience-ai/oneskills
GitHub	https://github.com/onescience-ai/OneScience	https://github.com/onescience-ai/oneskills

Citation and License

This repository is an independent engineering reproduction of the public CLM5-Emulator specifications.

The original paper is licensed under CC BY 4.0; the paper, CLM5, GSWP3, and FLUXNET-MTE data remain subject to their respective licenses and terms.