Framing the analysis
The central editorial decision was to measure exposure, rather than present a single risk score. A cell is exposed when its residents live within combustible vegetation or close to a large continuous fuel area. This is a physical condition that can be measured consistently across the country.
The map does not model ignition, fire behaviour, weather, wind, terrain slope, building conditions or local prevention measures. Those factors determine whether a fire starts, how it spreads and how severe its consequences may be. Keeping that distinction explicit prevented a structural-exposure map from being interpreted as a forecast or an individual assessment of danger.
Data and spatial model
The workflow joins the 100-by-100-metre population grid by Goerlich and Molla for Fundacion BBVA and Ivie, based on the 2021 census, with the most current Spanish Forest Map. The grid preserves the variation within municipalities that would be lost in municipal averages.
Combustible vegetation was identified from the forest map and dissolved into continuous areas. For the distance-based criterion, only areas of at least 500 hectares were retained, avoiding the classification of a settlement as exposed because of isolated patches of vegetation.
Each inhabited cell was then assigned one of two conditions:
- Intermix: at least half of the cell is covered by combustible vegetation.
- Interface: the cell is not intermix but sits within the selected distance of a continuous fuel area of at least 500 hectares.
The main scenario uses 400 metres. Sensitivity scenarios at 100 and 2,400 metres were also calculated to show how the total changes under a more restrictive direct-contact threshold and a broader urban-forest-interface reference.
To better approximate the current state of the available fuel, cells inside EFFIS/Copernicus fire perimeters from 2022 to 2026 were excluded. Recent fires can substantially alter vegetation, while areas burned further in the past may already have partly recovered.
From analysis to an interactive map
The spatial processing and aggregations were carried out in R. The pipeline prepares national fuel layers, identifies continuous large fuel areas, calculates cell-level exposure, aggregates results by municipality, province and autonomous community, and exports web-ready map sources.
This was my first project using MapLibre GL, an open-source map-rendering stack, together with PMTiles. The PMTiles protocol lets the browser request only the byte ranges needed for the visible map area, avoiding the cost of downloading national-scale source files in full.
National raw geometry would be too heavy for an interactive map, so the workflow exports separate tiled sources for population, fuel and exposure. The fuel layer is partitioned by autonomous community, dissolved by editorial fuel class and simplified before tiling. In contrast, the population and exposure layers preserve the geometry of their 100-metre cells without simplification, so that the grid remains faithful and rectangular as readers zoom in. This combination keeps the map responsive while allowing readers to switch between exposure classes, exposed population and the full population grid.
The visual hierarchy separates the conditions readers need to understand: combustible vegetation, population density, direct intermix and the three distance bands up to 400 metres. Local case studies and recent burned-area perimeters provide geographic context without changing the core measure.
Key decisions
- We used 100-metre cells as the base unit so the analysis could represent settlement patterns within municipal boundaries.
- We required fuel areas of at least 500 hectares for the interface criterion, focusing the distance calculation on continuous vegetation capable of sustaining a large fire.
- We used 400 metres as the main scenario because it is a defensible, operational threshold in regional regulations; the 100- and 2,400-metre scenarios make the choice transparent.
- We excluded recent burned perimeters to avoid treating recently altered vegetation as unchanged fuel.
- We described the result as exposure throughout the article and interface, rather than conflating it with fire probability or complete risk.
Result and limitations
The project turns several public spatial datasets into a national, explorable picture of where the urban fabric meets combustible vegetation. It helps identify places where prevention, planning and civil-protection measures could be prioritised by the people potentially protected, rather than by area alone.
The results remain an estimate. The population grid is modelled from the 2021 census, and neither the forest cover nor the built environment is static. The analysis also does not capture daily fire weather, evacuation capacity or the state of individual homes. These limitations are part of the result: the map shows a baseline of physical exposure that needs to be combined with local knowledge and other risk factors.