IberFire -- a detailed creation of a spatio-temporal dataset for wildfire risk assessment in Spain

TL;DR

IberFire delivers a high-resolution 1 km × 1 km × day datacube for wildfire risk in Spain, integrating 120 features across eight categories from open data sources to enable advanced ML/DL modelling and climate analyses. The paper details a reproducible workflow—from grid generation and feature derivation to data validation and cross-source consistency checks with AEMET—plus a practical demonstration using an XGBoost classifier that achieves AUROC 0.95 on 2024 data. Validation shows strong data correctness, thoughtful handling of missing values, and coherent fire-risk maps that align with historical fire patterns, underscoring IberFire’s utility for predictive modelling and policy support. The dataset and code are openly available to support open research, replication, and extension to other spatio-temporal environmental phenomena.

Abstract

Wildfires pose a threat to ecosystems, economies and public safety, particularly in Mediterranean regions such as Spain. Accurate predictive models require high-resolution spatio-temporal data to capture complex dynamics of environmental and human factors. To address the scarcity of fine-grained wildfire datasets in Spain, we introduce IberFire: a spatio-temporal dataset with 1 km x 1 km x 1-day resolution, covering mainland Spain and the Balearic Islands from December 2007 to December 2024. IberFire integrates 120 features across eight categories: auxiliary data, fire history, geography, topography, meteorology, vegetation indices, human activity and land cover. All features and processing rely on open-access data and tools, with a publicly available codebase ensuring transparency and applicability. IberFire offers enhanced spatial granularity and feature diversity compared to existing European datasets, and provides a reproducible framework. It supports advanced wildfire risk modelling via Machine Learning and Deep Learning, facilitates climate trend analysis, and informs fire prevention and land management strategies. The dataset is freely available on Zenodo to promote open research and collaboration.

Figures (8)

• Figure 1: Left: wildfires that occurred from 2008 to 2024, according to data from EFFIS. Right: selected area of interest to build the datacube.
• Figure 2: Left: Example of vectorial data (blue), some burned areas retrieved from EFFIS. Right: Example of raster data, elevation values on the same region as the left plot at a 1km$\times$1km resolution.
• Figure 3: Visual representation of a datacube.
• Figure 4: Left: raw geometries of the fire events downloaded from EFFIS. Right: the same geometries intersected with the spatial grid.
• Figure 5: Left: wind vectors that represent the sum of ERA5-Land u-wind and v-wind components. Right: the wind direction values of those same vectors.