orion-ai-lab/seasfire_monthly

---
license: cc-by-4.0
task_categories:
- time-series-forecasting
tags:
- wildfire
- earth-observation
- climate
- satellite
- seasonal-forecasting
- zarr
- xarray
size_categories:
- 10B<n<100B
language:
- en
---

# SeasFire monthly: Seasonal Fire Forecasting Datacube (Monthly Resolution)

## Dataset Description

The SeasFire monthly datacube is a comprehensive Earth observation dataset designed for **seasonal wildfire forecasting** using machine learning. It combines nearly 20 years (2001-2021) of multi-source satellite, meteorological, climatological, and human influence data into a unified, analysis-ready format.

This dataset is the result of extensive preprocessing and harmonization of the original [SeasFire v0.4 datacube](https://zenodo.org/records/7108392), with additional features including:
- Monthly temporal aggregation for seasonal forecasting tasks
- Integrated drought indices (SPEI at multiple timescales)
- Regional masks for targeted analysis (Greece, California, New South Wales)
- SATCLIP satellite image embeddings for enhanced feature representation

### Dataset Summary

- **Temporal Coverage**: 2001-2021 (20 years)
- **Temporal Resolution**: Monthly
- **Spatial Resolutions**: 0.25° (~25km)
- **Format**: Zarr (ZIP-compressed for portability)
- **Variables**: Refer to [SeasFire v0.4 datacube](https://zenodo.org/records/7108392) for comprehensive variable analysis.
- **Use Cases**: Wildfire prediction, seasonal forecasting, climate analysis, Earth system modeling

## Available Files

### Preprocessed Datacubes (Ready to Use)

| File | Size | Resolution | Description |
|------|------|------------|-------------|
| `seasfire_orora_v0.1.zip` | 15.8 GB | 0.25° spatial, monthly temporal | Main datacube with all features |
| `seasfire_orora_1deg_v0.1.zip` | 1.26 GB | 1° spatial, monthly temporal | Coarsened version for faster processing |

**These are the recommended files for most users.** They include monthly aggregation, drought indices, regional masks, and metadata.

### SATCLIP Embeddings (Optional Enhancement)

| File | Size | Resolution | Description |
|------|------|------------|-------------|
| `satclip-embedding.zip` | 3.78 GB | 0.25° | Full SATCLIP embeddings from satellite imagery |
| `satclip-embedding_1deg.zip` | 243 MB | 1° | Coarsened SATCLIP embeddings |
| `satclip_pcs_v0.1.zip` | 163 MB | 0.25° | Principal components (first 5) of SATCLIP embeddings |

**SATCLIP embeddings** are learned representations from satellite images using contrastive learning. They can enhance model performance but are optional.

### Original SeasFire Cube

| File | Size | Resolution | Description |
|------|------|------------|-------------|
| `original_cube/seasfire_v0.4.zip` | 43.9 GB | 0.25° spatial, 8-daily temporal | Original SeasFire datacube before preprocessing |
| `original_cube/seasfire_1deg_v0.4.zip` | 2.69 GB | 1° spatial, 8-daily temporal | Coarsened original datacube |

**Use these only if** you need the original 8-daily temporal resolution or want to customize the preprocessing pipeline.

## Additional Data Variables

### Drought Indices (SPEI)
- **spei_1, spei_3, spei_6, spei_12, spei_24, spei_36, spei_48**: Standardized Precipitation-Evapotranspiration Index at multiple timescales (1-48 months)

### SATCLIP Features (in separate files)
- **satclip_embeddings**: Learned representations from satellite imagery
- **satclip_embeddings_pc1 to pc5**: Principal components of embeddings

## Loading the Data

### Python with xarray

```python
import xarray as xr

# Load the main datacube
ds = xr.open_zarr("seasfire_orora_v0.1.zip", consolidated=True)

# Or load the 1-degree version for faster processing
ds = xr.open_zarr("seasfire_orora_1deg_v0.1.zip", consolidated=True)

# Explore the dataset
print(ds)
print(ds.data_vars)

# Select a specific time period
ds_subset = ds.sel(time=slice("2020-01-01", "2021-12-31"))

# Select a specific region (e.g., California)
california_data = ds.where(ds.regions_of_interest == 2, drop=False)
```

### Downloading with Hugging Face Hub

```python
from huggingface_hub import hf_hub_download

# Download a specific file
file_path = hf_hub_download(
repo_id="orion-ai-lab/seasfire_orora",
filename="seasfire_orora_v0.1.zip",
repo_type="dataset"
)
```

### Using the Download Script

```bash
# Clone the repository
git clone https://github.com/Orion-AI-Lab/orora-deliverable-ml-lab.git
cd orora-deliverable-ml-lab

# Install dependencies
pip install -r requirements.txt

# Download the dataset
python datacube/huggingface/dataset_download.py \
./local_data \
--repo_id orion-ai-lab/seasfire_orora
```

## License

This dataset is released under the **Creative Commons Attribution 4.0 International (CC BY 4.0)** license.

You are free to:
- Share: copy and redistribute the material
- Adapt: remix, transform, and build upon the material

Under the following terms:
- Attribution: You must give appropriate credit and indicate if changes were made

## Contact & Support

- **Repository**: **Private**
- **Organization**: [Orion-AI-Lab](https://github.com/Orion-AI-Lab)

## Version History

- **v0.1** (2024): Initial release
- Monthly temporal aggregation
- Integrated drought indices (SPEI)
- Regional masks for Greece, California, and New South Wales
- SATCLIP embeddings (optional)
- Multiple spatial resolutions (0.25° and 1°)

---

**Dataset Status**: Active | **Last Updated**: February 2026