Wildfire — annualised loss (current)
How often fire burns a place, multiplied by how much damage the local fire weather implies.
Source: MODIS/Fire_cci burned-area climatology × design Fire Weather Index (NASA NEX-GDDP-CMIP6) Year: current conditions Category: Physical risks · Direct Physical Loss engine inputs Coverage: Global Format: Raster grid (~25 km) Used in risk analysis: Yes — gates Mitigating services (Fire risk) at ≥ 0.01 %/yr
What it shows
How much of a building's value wildfire is expected to destroy in an average year, as a percentage per year — an average annual loss. It multiplies the two things that make fire costly: how often the land actually burns, and how destructive the local fire weather makes a fire once it starts.
Neither factor is honest on its own. The Fire Weather Index, the meteorological measure of how readily a fire spreads, puts the Sahara at the top of its scale for a hazard whose burn probability there is near zero; the burned-area record on its own reads near-empty across the Mediterranean. Only the product is reported.
How it is built
Pixel by pixel, the annual burn probability — the observed burned-area fraction from the ESA Fire_cci / C3S OLCI climatology — is multiplied by the loss rate a published damage function assigns to the local fire weather. The design Fire Weather Index is converted to fireline intensity through the Van Wagner relation, and the damage function is read on that intensity.
How to read it
Higher values mean more expected loss per year. The figure is not a given site's loss: it assumes a building at the fuel edge, the most exposed position there is. A site's own exposure is applied separately, on the activity side. Read the layer as a ranking of fire country, not as a price.
Class thresholds
A site is flagged on Mitigating services when the annualised loss reaches 0.01 %/yr.
Because the value is the loss of a hypothetical building, "0.01 % of asset value" is not this site's expected loss and must not be read as one. The threshold is a calibration point, not a materiality figure.
How this threshold was calibrated
Against Fire Weather Risk 2050 (France, DRIAS), Météo-France's independent reference for France, on the question a flag actually has to answer: does it land where the reference puts its fire hotspots — the worst 10 % of its cells? Agreement peaks between 0.005 and 0.01 %/yr and falls away by 0.05.
For audit: at the chosen 0.01 %/yr the flag covers 14.7 % of France and finds 55 % of those hotspots. That share is measured at the operating threshold, which casts a wider net than a decile — it is not comparable with the equal-coverage figures quoted below, which restrict every candidate to its own top 10 %.
Why this layer
It replaces Wildfire Hazard (NASA FIRMS NOAA-21), which counts satellite detections of active fires and so inflates wherever crop residue is burned every year.
Compared at equal coverage — each layer restricted to its own top 10 % of French cells, so both flag the same area — this layer finds 39.9 % of the DRIAS hotspots where FIRMS finds 20.2 %, against the 10 % a random draw would find. Region by region, FIRMS under-flags California (25.8 % against 86.4 % here), the Cerrado (59.5 against 99.7) and boreal Siberia (5.6 against 22.1), while over-flagging north-west Europe (5.1 against 1.2).