Book a call
Insurance // Label type 06

Wildfire burn area labeling for insurance

Burn area labeling outlines the land a wildfire burned and, for insurance, the settlements and buildings at or inside its edge. It gives wildfire risk and claims models an observed footprint instead of a reported one.

Landsat image of the Balmoral fire near Sydney: a large reddish-brown burned area below farmland and small towns.
RawLabeled
  • Burned area1 polygon
  • Built-up exposure4 polygons
  • Water1 polygon
Example: Balmoral fire south-west of Sydney, 5 November 2013, Landsat 8 OLI (NASA Earth Observatory / USGS). Labels drawn by IKT Spīdolas for illustration.

What burn areas labeling is

Burned land is clear in the shortwave infrared bands of Sentinel-2 and Landsat, even through smoke. We trace the fire perimeter and the unburned islands inside it, then add the exposure layer that matters to an insurer: built-up areas and buildings at the edge or inside the burn. The example is the 2013 Balmoral fire south-west of Sydney.

How we label it

Footprints are labeled on post-fire imagery compared to a pre-fire scene, so cleared land and shadows are not mistaken for burn. Burn severity classes are available when the model needs more than burned / unburned.

Who uses burn areas labels

  • Claims triage

    Identify which policies sit inside or next to a burn before inspection.

  • Wildfire risk models

    Historic footprints at the edge of towns, where losses happen, to train and test risk scores.

  • Underwriting

    Past fire activity around a location as an input to pricing.

  • Portfolio exposure

    Sum insured inside a footprint within days of a fire.

How a project runs

Four steps, each ending in something you can check before the next one starts.

  1. Brief

    Area, dates and class list agreed in writing, with example images for every class and its edge cases.

    You getWritten labeling guideline

  2. Imagery

    Your imagery, or sourced for you: free Sentinel and Landsat, or commercial high-resolution scenes.

    You getScene list with dates and resolution

  3. Labeling

    Annotators work to the guideline. Anything the imagery cannot settle is flagged, not guessed.

    You getLabels plus a log of flagged cases

  4. Check & delivery

    Every class is reviewed against a sample before the data leaves us.

    You getData and per-class accuracy, in your format and CRS

Questions about burn areas

What is the difference from burn scars?

The same fire, a different question. Burn areas for insurance focus on the footprint against buildings; burn scars for forestry focus on the forest that burned and how it recovers.

How soon after a fire can you label?

As soon as there is a clear post-fire image. Sentinel-2 revisits every few days.

Can you show which buildings burned?

That needs high-resolution imagery. See building damage labeling.

Related label types

Next step

Need burn areas labeled?

Tell us your area, your imagery and what your model has to learn. We reply with a proposed class list and a quote.