Sentinel-2 Wildfire Risk Classification Dataset for the Iberian Peninsula [Dataset]

Abstract

[EN]This dataset was developed to support the development, training and evaluation of artificial intelligence and machine learning models for wildfire risk prediction in the Iberian Peninsula.

The dataset integrates satellite observations from Sentinel-2 with meteorological information from ERA5 and fire-weather indicators derived from the Fire Weather Index (FWI) system. Data were collected for the period from 16 July 2024 to 31 August 2024, with observations obtained approximately at midday following the temporal criteria established by the Spanish State Meteorological Agency (AEMET).

Sentinel-2 data include ten spectral bands (B2, B3, B4, B5, B6, B7, B8, B8A, B11 and B12).

ERA5 meteorological information includes 2-metre temperature, 2-metre dewpoint temperature, the U and V components of 10-metre wind, and total precipitation. These variables were used to calculate the components of the Fire Weather Index using the FWIfunctions library (version 0.1.10), including Fine Fuel Moisture Code (FFMC), Duff Moisture Code (DMC), Drought Code (DC), Initial Spread Index (ISI), Buildup Index (BUI), and the resulting Fire Weather Index (FWI).

The continuous FWI values were subsequently classified into five wildfire-risk categories: low, medium, high, very high and extreme. The class thresholds were established using the 40th, 65th, 85th and 95th percentiles of the FWI distribution, respectively.

A common spatial grid with a resolution of 0.1° × 0.1° was used to harmonize Sentinel-2 and ERA5 information. The study area was defined using the bounding box [44, -10, 36, 4]. Each record represents a spatiotemporal observation linking satellite-derived information with meteorological and FWI-derived wildfire-risk information for a given location and date.