# ClimSim Low-Resolution Expanded Inputs (ClimSim-Online)

## Facts

- Provider: LEAP NSF Science and Technology Center, Columbia University, and ClimSim consortium
- License: CC-BY-4.0 (https://github.com/leap-stc/ClimSim)
- Size: 100,000 files
- Format: NetCDF (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/LEAP/ClimSim_low-res-expanded
- Access: public, readable without an account
- Category: Climate & Weather
- Source: https://arxiv.org/abs/2306.08754
- Page: https://scigantic.com/archive/dee0066f-b3a7-5fa0-aaa6-1f2fed0e932b
- Keywords: Climate & Weather, climate model, NetCDF, Hugging Face

## About

ClimSim, from the LEAP NSF Science and Technology Center and partners, is the largest dataset built for hybrid ML-physics climate emulation: input and output vectors from the multi-scale climate model E3SM-MMF that isolate how embedded high-resolution cloud-resolving physics changes the host model's macro-scale state, so neural networks can learn to replace the expensive subgrid physics. Data were saved every 20 minutes for 10 simulated years as paired netCDF files, E3SM-MMF.mli.* (inputs) and E3SM-MMF.mlo.* (targets), with scalar and vertically resolved variables on ncol and lev. This repository expands ClimSim_low-res for the ClimSim-Online framework: each .mlexpand. file holds the same variables as the matching .mli. input file plus additional features, among them large-scale dynamical forcing, convection memory from the previous step and the cosine and sine of latitude, described in Section 6.3.3 of the supplementary information of the ClimSim-Online paper. These inputs support hybrid E3SM-MMF-NN runs where the emulator is coupled back into the host model and tested online. The listing (measured 2026-11-14) shows a train/ folder of monthly subfolders; the repository uses about 1.5 TB. Open with xarray or netCDF4 alongside ClimSim_low-res targets. Data are released under CC-BY-4.0 (code Apache-2.0); cite arXiv:2306.08754.

## Load the data

```python
from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id="LEAP/ClimSim_low-res-expanded", filename="train/0001-02/E3SM-MMF.mlexpand.0001-02-01-02400.nc", repo_type="dataset")
print(path)
```

## Data structure

File types in sampled files: NetCDF

Example file: `train/0001-02/E3SM-MMF.mlexpand.0001-02-01-02400.nc`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/dee0066f-b3a7-5fa0-aaa6-1f2fed0e932b
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/dee0066f-b3a7-5fa0-aaa6-1f2fed0e932b
- Over MCP: `get_schema_card` with id `dee0066f-b3a7-5fa0-aaa6-1f2fed0e932b` at https://api.scigantic.com/mcp

## Related topics

- [Climate & Weather datasets](https://scigantic.com/topics/climate-and-weather)
- [NetCDF datasets](https://scigantic.com/topics/netcdf)
- [Climate model datasets](https://scigantic.com/topics/climate-model)
