# ImpactMesh-Fire: Multimodal Multitemporal Wildfire Burn Scar Dataset

## Facts

- Provider: IBM Research, DLR, ESA Phi-lab
- License: CC-BY-4.0 (https://huggingface.co/datasets/ibm-esa-geospatial/ImpactMesh-Fire)
- Size: 22 files
- Stored at: Hugging Face, https://huggingface.co/datasets/ibm-esa-geospatial/ImpactMesh-Fire
- Access: public, readable without an account
- Category: Earth Observation & Geospatial
- Source: https://github.com/IBM/ImpactMesh
- Page: https://scigantic.com/archive/11a284ee-90d1-548e-9a98-9693fd8841b9
- Keywords: Earth Observation & Geospatial, image segmentation, satellite imagery, time series, benchmark, Hugging Face

## About

Wildfire subset of ImpactMesh, the multimodal, multitemporal disaster mapping benchmark from IBM, DLR and ESA Phi-lab. It pairs Sentinel-2 L2A (all bands), Sentinel-1 RTC SAR and Copernicus DEM with burn scar annotations from the Copernicus Emergency Management Service, with four time steps per sample (pre-month, pre-event, event and post-event). Version 1 holds 16,512 samples covering 229 wildfire events (11,852 train, 1,864 validation, 2,796 test), cleaned to reduce label noise, and a 1,270-sample test_holdout of events absent from training and validation. The listing (measured 2026-11-14) shows train/, val/, test/ and split/ folders with per-modality tar archives (S2L2A, S2RGB, S1RTC, DEM, MASK); the repository uses about 153 GB. Untar the archives as the card shows and read samples with rasterio or xarray. Intended for burned area segmentation, fire impact mapping and fine-tuning geospatial foundation models with optical and radar time series. Released under CC-BY-4.0, with Copernicus Sentinel attribution; a companion flood subset is ImpactMesh-Flood.

## Load the data

```python
from huggingface_hub import hf_hub_download
path = hf_hub_download(repo_id="ibm-esa-geospatial/ImpactMesh-Fire", filename="test/DEM.tar", repo_type="dataset")
print(path)
```

## Data structure

Example file: `test/DEM.tar`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/11a284ee-90d1-548e-9a98-9693fd8841b9
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/11a284ee-90d1-548e-9a98-9693fd8841b9
- Over MCP: `get_schema_card` with id `11a284ee-90d1-548e-9a98-9693fd8841b9` at https://api.scigantic.com/mcp

## Related topics

- [Earth Observation & Geospatial datasets](https://scigantic.com/topics/earth-observation-and-geospatial)
- [Satellite imagery datasets](https://scigantic.com/topics/satellite-imagery)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [Image segmentation datasets](https://scigantic.com/topics/image-segmentation)
- [Time series datasets](https://scigantic.com/topics/time-series)
