European Multi-Hazard Dataset in Sparse Array Format

This dataset contains information on individual and multi-hazard events in Europe. The individual hazard data cover heatwaves, wildfires, floods, and windstorms, while the multi-hazard data describe pairs of events that overlap in both space and time, specifically heatwave–wildfire and windstorm–flood events. The data are provided in sparse array NetCDF files and .csv files, with the .csv files containing information on the identified hazard events and their characteristics. The data are further described below Single hazard dataset description This dataset provides pre-processed hazard data for four natural hazards across Europe: Heatwaves Wildfires Windstorms Floods The data have been harmonized and converted into a sparse array format, designed to improve storage efficiency and facilitate the handling of large-scale spatial hazard datasets. The dataset is distributed in NetCDF (.nc) format. Each hazard is provided as a separate NetCDF file, with the hazard data and associated variables organized within the file structure. Original data sources The hazard datasets were derived from open-source datasets and subsequently pre-processed to the sparse array format. Hazard Original data citation Variable Heatwaves Copernicus Climate Change Service. (2019). ERA5-Land hourly data from 1950 to present [Data set]. doi:10.24381/CDS.E2161BAC Temperature of air at 2m above the surface of land, sea or in-land waters Wildfires San-Miguel-Ayanz, J., Schulte, E., Schmuck, G., Camia, A., Strobl, P., Liberta, G., Giovando, C., Boca, R., Sedano, F., Kempeneers, P., McInerney, D., Withmore, C., de Oliveira, S. S., Rodrigues, M., Durrant, T., Corti, P., Oehler, F., Vilar, L., & Amatulli, G. (2012). Comprehensive Monitoring of Wildfires in Europe: The European Forest Fire Information System (EFFIS). In Approaches to Managing Disaster - Assessing Hazards, Emergencies and Disaster Impacts. InTech. https://doi.org/10.5772/28441 Burnt area mapped using Sentil2/MODIS images Windstorms Copernicus C3S. (2025). Windstorm tracks and footprints derived from reanalysis over Europe between 1940 to present [Dataset]. ECMWF. https://doi.org/10.24381/BF1F06A9 maximum 10m wind gust over a 72-hour time window Floods Betterle, A., & Salamon, P. (2025). Satellite-derived flood depth maps for Europe [Dataset]. European Commission, Joint Research Centre (JRC). https://doi.org/10.2905/0BC96690-B89C-4909-9166-C2C322A20130 flood depth Please refer to the original data providers and associated publications for detailed information on the generation and methodology of the source datasets. Users of this dataset should also cite the original data sources where appropriate. Data format and structure The hazard data consists of four NetCDF (.nc) files, each containing hazard events for one hazard type. Each hazard is identified using a hazard-specific acronym, which is also used in the naming of the event groups and hazard value variables Hazard type Acronym NetCDF file Event group naming Unit Start year End year Floods fl flevents.nc fl<event_id> (e.g., fl10 cm 2015 2024 Windstorms ws wsevents.nc ws<event_id> (e.g., ws10) m/s 2015 2024 Wildfires wf wfevents.nc wf<event_id> (e.g., wf10) Binary indicator (0 = not burned; 1 = burned) 2008 2026 Heatwaves hw hwevent.nc hw<event_id> (e.g., hw10) °C 2008 2025 Each NetCDF file contains multiple groups, with each group representing an individual hazard event. Event groups are named using the hazard-specific acronym followed by the corresponding event identifier. For example, fl10 represents flood event 10, while ws100 represents windstorm event 100. Event-level data attributes Each event group contains spatial, temporal, identification, and hazard intensity information. Variable Description Units/format Availability coords Sparse spatial coordinates identifying the locations affected by the hazard event. Index coordinates All hazards <acronym>values The hazard values associated with the affected locations, for example flvalues for flood Hazard specific All hazards start_date Date on which the event started. For heatwaves, this variable contains the dates corresponding to each day of the heatwave event. Days since 1970-01-01 All hazards end_date Date on which the event ended. Days since 1970-01-01 fl, ws, wf event_id Unique identifier for the hazard event Integer All hazards latitude Latitude associated with the spatial domain Degrees north (°N) All hazards longitude Longitude associated with the spatial domain Degrees east (°E) All hazards Example event file structure A flood event group is organized as follows: flevents.nc│├── fl10│ ├── coords│ ├── flvalues│ ├── start_date│ ├── end_date│ ├── event_id│ ├── latitude│ └── longitude│├── fl11│ ├── coords│ ├── flvalues│ ├── start_date│ ├── end_date│ ├── event_id│ ├── latitude│ └── longitude│└── ... The same general structure is used for wildfire, heatwave, and windstorm files, using the corresponding hazard acronyms and hazard value variables. Sparse array representation The dataset uses a sparse array representation to efficiently store hazard information. Instead of storing values for the complete spatial domain, the dataset stores information only for locations associated with a hazard event. The coords variable identifies the positions of relevant values within the spatial domain, while the corresponding hazard-specific values variable (flvalues, wfvalues, hwvalues, or wsvalues) stores the hazard magnitude at those locations. Together, these variables allow the spatial representation of each event to be reconstructed. For floods, wildfires, and windstorms, the sparse coordinate structure is two-dimensional. Heatwaves use a three-dimensional sparse representation that additionally includes a temporal dimension, allowing the spatial extent and hazard values to vary throughout the duration of an event. Single hazard information .csv files In addition to the NetCDF files, each single hazard file is accompanied by a .csv with hazard information, such as the group names. The csv’s are structured as follows Variable Description hazardtype The hazard acronym groupname The group name in the respective .nc file start_date Date on which the event started end_date Date on which the event ended minlon Minimum longitude of the event maxlon Maximum longitude of the event minlat Minimum latitude of the event maxlat Maximum latitude of the event intensity The intensity of the hazard event Intensity unit The intensity unit of the hazard intensity Multi-hazard information .csv files The multi-hazard files (wf_hw.csv and fl_ws.csv) contain pairs of hazard events that overlap in both space and time. These overlaps were identified using an updated version of the MYRIAD-HESA algorithm (Claassen et al., 2023). The variables groupname, start_date, end_date, hazardtype, minlon, maxlon, minlat, maxlat, intensity, and intensity unit are provided for each hazard event as described above. The suffixes _1 and _2 indicate the first and second hazard in each overlapping pair. The column names in the multi-hazard .csv files are therefore: groupname_1, groupname_2, start_date_1, end_date_1, start_date_2, end_date_2, hazardtype_1, minlon_1, maxlon_1, minlat_1, maxlat_1, intensity_1, intensity unit_1, hazardtype_2, minlon_2, maxlon_2, minlat_2, maxlat_2, intensity_2, intensity unit_2. Usage and reproducibility Additional documentation, and code for reading and using the dataset will be made available in a dedicated GitHub repository. The repository will include: Code for loading and interpreting the NetCDF files Examples demonstrating how to reconstruct and work with the sparse array representation Algorithm to identify multi-hazard events The GitHub repository and associated code will be shared soon.