How a burnt area is produced
EFFIS maps the burnt area after the event from satellite imagery and attaches an attribute set to every outline: fire date, end date, last update, country, province and municipality, area in hectares, and the shares of the affected land cover classes, broadleaved, coniferous and mixed forest, sclerophyllous and transitional vegetation, agricultural and artificial surfaces and others. One field additionally states the share lying inside a Natura 2000 site.
It is precisely from these shares that the emission calculation draws its vegetation types: without the mapped land cover share it would have to guess a combustion factor.
The minimum area
The layer name still carries the word MODIS, but since 2018 the polygons come predominantly from Sentinel-2 at twenty metres. The frequently quoted minimum of 30 hectares applies only to the MODIS trigger and is simply wrong for today's data.
Measured across the working area for 2026: the median is 12 hectares, the minimum 1 hectare, and 26 polygons carry 0 hectares because areas are rounded to whole hectares. The practical lower bound therefore lies at roughly one hectare.
Small, quickly extinguished fires are nevertheless missing, systematically so. Anyone inferring the workload of a fire station from this statistic will underestimate it. For area and emission balance it matters far less: the totals are carried by the large events.
Coverage outside the European Union
EFFIS is not restricted to EU member states. Polygons with a full attribute set exist for Bosnia and Herzegovina, Serbia, Montenegro, North Macedonia and Kosovo as well, the figures for the current season are below.
The restriction lies elsewhere: the published completeness figure of roughly 95 per cent of burnt area applies explicitly to the European Union only. For the other countries of the working area no figure is published. samoEKO therefore presents this data as satellite-based recording and never as an official complete survey.
A second trap concerns headline figures: the tabular annual statistic from EFFIS contains Serbia and North Macedonia but not Bosnia and Herzegovina, Montenegro or Kosovo. For working-area figures samoEKO therefore aggregates the polygons itself and never adopts the ready-made table.
| Country | Polygons | Area |
|---|---|---|
| Bosnia and Herzegovina | 466 | 21,833 ha |
| Montenegro | 323 | 11,836 ha |
| Croatia | 96 | 3,866 ha |
| Serbia | 185 | 3,674 ha |
| North Macedonia | 95 | 2,686 ha |
| Kosovo | 72 | 1,425 ha |
| Slovenia | 6 | 387 ha |
Own count of the EFFIS polygons as of 2 August 2026. Bosnia and Herzegovina is thus the most affected country in the working area.
Burn severity (dNBR): open
How severely an area has burnt is not stated by its outline. Between a surface fire the stand survives and a crown fire that destroys it lie, at identical area, entirely different consequences, for reforestation, for erosion control and for the amount emitted.
The usual measure is the differenced normalized burn ratio. From the near and short-wave infrared of a scene a ratio is formed per pixel that responds sensitively to burnt vegetation; the difference between a pre-fire and a post-fire scene gives the severity, which is then classified.
In samoEKO this route is prepared but not finished. Scene selection and band access exist through the Sentinel-2 chain, the near and short-wave infrared are carried for every selected scene. What is missing is the cloud mask, the calculation and the class boundaries. Until then the EFFIS polygons are the robust post-fire product. This gap is stated here because a half-finished severity classification in a situation picture would be worse than none.