---
# For reference on dataset card metadata, see the spec: https://github.com/huggingface/hub-docs/blob/main/datasetcard.md?plain=1
# Doc / guide: https://huggingface.co/docs/hub/datasets-cards
{}
---

# Dataset Card for ERA5 climate dataset


<!-- Provide a quick summary of the dataset. -->

This dataset is for reproducing [ClimODE](https://github.com/Aalto-QuML/ClimODE) results, which aims to model the climate efficiently with physics-inspired neural networks. It contains a 2004-2018 subset of the [weatherbench](https://dataserv.ub.tum.de/index.php/s/m1524895) data at 5.625deg granularity. 

Made with :heart: from by students at [USI](https://www.usi.ch/it).

## Dataset Details

### Dataset Sources [optional]

<!-- Provide the basic links for the dataset. -->

- **Repository:** https://dataserv.ub.tum.de/index.php/s/m1524895
- **Paper** https://cds.climate.copernicus.eu/datasets/reanalysis-era5-single-levels

## Uses

Climate modeling and prediction

### Direct Use

<!-- This section describes suitable use cases for the dataset. -->

This dataset is used as a supporting resource in [Climode-reproducibility](https://github.com/tornikeo/climode-reproducibility) repository, which aims to reproduce and validate the original paper authors' claims. 

You can download the dataset like so:

```py
from huggingface_hub import snapshot_download

if __name__ == '__main__':
    snapshot_download(repo_id="TornikeO/era5-5.625deg",
                    repo_type='dataset',
                    local_dir='climode/ClimODE/era5_data', 
                    allow_patterns="*")
```

## Dataset Structure

<!-- This section provides a description of the dataset fields, and additional information about the dataset structure such as criteria used to create the splits, relationships between data points, etc. -->

The dataset is a set of `.nc` files. These can be read as follows:

```py
import xarray as xr

# Download the dataset from TornikeO/era5-5.625deg

data = xr.open_mfdataset(
    'era5-5.625deg/geopotential_500/*.nc', 
    combine='by_coords')
data = data.resample(time="6H").nearest(tolerance="1H") # Setting data to be 6-hour cycles
data_global = data.sel(time=slice('2006','2018')).load() # select subset only
```

## Dataset Creation


### Source Data

<!-- This section describes the source data (e.g. news text and headlines, social media posts, translated sentences, ...). -->

This dataset was generated by downloading [WeatherBench](https://dataserv.ub.tum.de/index.php/s/m1524895) data
#### Data Collection and Processing

<!-- This section describes the data collection and processing process such as data selection criteria, filtering and normalization methods, tools and libraries used, etc. -->

Dataset was downloaded and processed using python, with a simple , using [automated python download script](https://github.com/tornikeo/climode-reproducibility/blob/main/scripts/download.py).

## Citation [optional]

<!-- If there is a paper or blog post introducing the dataset, the APA and Bibtex information for that should go in this section. -->

**BibTeX:**

[More Information Needed]

**APA:**

[More Information Needed]

## Glossary [optional]

<!-- If relevant, include terms and calculations in this section that can help readers understand the dataset or dataset card. -->

[More Information Needed]

## More Information [optional]

[More Information Needed]

## Dataset Card Authors [optional]

The Sole author is [Tornike Onoprishvili](https://github.com/tornikeo/).

## Dataset Card Contact

[More Information Needed]