What wildfires cost

A hotspot on a map says very little. What speaks is the area gone up in smoke, the carbon released, the air we breathe and the bill. Here are those four figures, and exactly how they are obtained.

Three confidence levels, never mixed

The hectares are measured: EFFIS maps the actual perimeters. The CO₂ and the particles are modelled by Copernicus. The cost is estimated by us, and it is the only figure we alone are responsible for. Every panel on this site carries the matching label.

Year 2026

What that means

2026

96,591

hectares burned

measured · EFFIS

kilotonnes de CO₂

modelled · Copernicus · as much as 6,105 cars for a year

tonnes de particules fines

modelled · PM2.5, the ones that reach the lungs

≈ 650

millions d'euros

estimated · insured damage and aerial firefighting, prorated from 2022

The hectares are a measurement: EFFIS maps the actual perimeters of fires above 30 hectares. The CO₂ and the particles come from the Copernicus model. The euros are a rule of three on 2022, the only year with figures: they count insured damage and the extra cost of aerial firefighting, not ground operations nor uninsured forest. The detail behind each figure

Where each figure comes from

Hectares burned, measured

EFFIS (Copernicus) maps the actual perimeters of fires from satellite imagery. This is not a calculation but a survey. Limit: only fires above 30 hectares are mapped, so the total is a floor. The complete official French measurement is BDIFF, published a year behind.

Why our figure differs from the one in the papers. During the season, the Interior Ministry regularly announces higher national totals. We do not know precisely how they are built, so we do not repeat them. What we can say about our figure: it comes from EFFIS, it ignores fires below 30 hectares, and it is revised upwards for several days after a fire, as the imagery is processed. On 22 July 2026, EFFIS showed 42,810 hectares for the year; five days later, the same date had been corrected to 77,696. In the middle of the season, our total is therefore a floor catching up, not a live counter.

The complete French measurement remains BDIFF, published a year behind, and it does not match EFFIS either: depending on the year, it is above or below. We never add them together and we never present them as interchangeable.

CO₂ and fine particles, modelled

Copernicus derives these quantities from the power released by fires, measured by satellite, through its GFAS chain. These are orders of magnitude produced and validated by the European service, not by us.

Euros, estimated by a rule of three

Nobody publishes the bill for a season in progress. Only one year has its own: 2022, with about 72,000 hectares burned. From it we derive two ratios, which we apply to the measured hectares of the current year.

Insured damage: 6,250 € per hectare: Over 450 million euros in claims in 2022. Counts only what was insured, and forest almost never is.
Aerial firefighting: 458 € per hectare: 33 million euros in extra spending in 2022, from flight hours and water drops.

This really is the season's extra cost, not the budget for aerial assets, which runs at around 180 million a year. The distinction drives the method: a budget is paid whether it burns or not, so dividing it by the hectare would give a unit cost that falls in the worst years. The extra cost follows the season.

Left out: the ground operations of the fire services, for which no breakdown by fire type is published, uninsured forest and everything that is not compensated. The total is therefore a floor, like the hectares. We round it to two figures: a rule of three does not have the precision of a euro.

Sources: France Assureurs, natural events report 2022; French Senate, report on the civil protection water bomber fleet (r22-838); burned areas from the statistical service of the Ministry for the Ecological Transition (62,000 ha of forest and 10,000 ha of other vegetation).

Casualties and buildings, declared

It is BDIFF, the Agriculture Ministry's database, that records them, fire by fire. Neither the satellites nor EFFIS count casualties: nobody else publishes this figure.

11: deaths declared between 2006 and 2025
306: buildings completely destroyed over the same period

These are deaths declared under a vegetation fire, not all fire-related deaths. For scale: the Prométhée database counts 144 deaths in the Mediterranean zone between 1973 and 2021, about 70 % of them firefighters, and the deadliest fire remains the one at Cestas, in Gironde, in 1949, with 82 deaths.

The method we no longer use for hectares

This site first estimated burned areas from radiative power, for want of anything better. The range spanned a factor of two. Since the EFFIS perimeters were integrated, that chain only serves to understand how a satellite signal becomes tonnes, and to cross-check orders of magnitude.

1

Energy released

sum of measured radiative power × 6 h of observation

A satellite sees the fire for only a few seconds, twice a day. This is the most fragile assumption in the chain.
2

Biomass consumed

0.368 kg per megajoule

Relationship established by Wooster et al. (2005), used by the European Copernicus service.
3

CO₂

1.6 kg of CO₂ per kg of dry matter burned

Temperate forest emission factor (Andreae, 2019).
4

PM2.5 fine particles

15.5 g per kg of dry matter burned

Same source. These are the particles that reach deepest into the lungs.
5

Burned area

biomass ÷ fuel load, from 20 to 35 t/ha

Range calibrated on 2022, a year whose burned area is known (about 72,000 hectares). The official measurement will come from BDIFF, published a year behind.

The control point

A method that is never checked against anything is worth nothing. This one is calibrated on 2022, a year whose burned area in France is known independently: about 72,000 hectares.

Applied to 2022, the method gives 86,244 to 150,928 hectares. The known value, 72,000, falls outside: the method needs revisiting, and this finding stays on display.

This is what fixes the fuel load used. Without this control point, the area range would be twice as wide.

Year by year, measured

Areas mapped by EFFIS since 2006.

Year	Hectares burned	Mapped fires
2026 (in progress)	96,926	360
2025	36,951	257
2024	12,268	156
2023	22,350	219
2022	66,392	291
2021	30,652	214
2020	14,547	133
2019	43,602	304
2018	2,581	21
2017	20,626	91
2016	10,767	27
2015	2,046	15
2014	4,667	26
2013	891	9
2012	3,298	19
2011	2,205	22
2010	4,653	9
2009	7,974	16
2008	1,694	8
2007	2,631	12
2006	1,751	12

The sources, the licences and the freshness of the data are on the Data and licences page, and what each sensor measures is explained in the methodology.