orion-ai-lab/seasfire_monthly

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, 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 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

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

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

# 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

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