Wildfire Dataset from Satellite and Social Media Sources

This dataset bundle contains multi-source data used for near real-time wildfire detection and monitoring in the Attica region of Greece, as part of the EU-funded BeOpen project. It combines Earth Observation (EO) information with social media activity and fire danger indicators to support early warning, analysis, and response in an operational urban resilience context.

The full dataset collection includes:

- Historical Fires Dataset: Wildfire burned area scars that occurred between 2019 and 2021 in the Attica region, providing ground-truth information for validating fire detection and risk models. - Social Media Events Related to Fires: Clusters of geolocated social media events linked to identified fire incidents in Attica from 2019 to 2021, derived using anomaly detection on fire-related posts. - Social Media Posts Related to Fires: Fire-related posts from the X platform for 2019–2021 in Attica, including post identifiers, URLs, timestamps, locations extracted from text, links to related images, fire danger level, and relevance scores generated by CERTH processing. - Predicted Fire Danger Levels Using EFFIS Data: Predicted fire danger levels for areas of Athens, produced by an MLP model trained on SatCLIP-based embeddings and using fire danger information from the European Forest Fire Information System (EFFIS). - Copernicus/EO Air Quality Dataset: Sentinel-5 NO₂ air quality anomaly measurements for the Attica region between 2019 and 2021, used to support detection and characterization of wildfire-related events.

All datasets have been processed, harmonized, and managed through the BeOpen Framework, which provides storage (e.g., MinIO), cataloguing and API access (BeOpen CKAN federated with IDRA), data model mapping to standardized formats, and metadata quality validation to ensure reliable, interoperable use. Together, these resources enable the Near Real-Time Fire Event Detection Digital Service to fuse EO and non-EO data streams, identify abnormal patterns associated with wildfires, and expose results through interactive maps and dashboards for decision makers and emergency responders.

The datasets are publicly accessible via the European Data Portal and related endpoints associated with the BeOpen project.

Further information regarding our dataset and how to access it can be found here.