Skip to content

Control lines that hold only where the fire is weak? #254

Open

Description

@HugoFara

@filippi, would you want ForeFire to have control lines that hold only where the fire is weak? Roads and DFCI tracks often stop a flank or a night front, but not a head run.

I tried roads as non-burnable cells on a real case (Gonfaron, Var, 2021; ensemble as in #172), and a static break is wrong either way:

  • DFCI tracks as breaks stop the first run in the plain: the median head error goes from 4.9 km to 11.6 km. The real fire crossed every track at 3–7 km/h, with spot fires 1–2 km ahead.
  • Without them, the flanks keep going. The real head stopped on the D98 and the A92 track around midnight, with crews and tactical fire, and the flanks were held overnight. The ensemble puts p > 0.5 on 54 km² that never burnt, almost all of it on one flank.

Two ways I can see:

  1. A control-line layer: a node on a marked cell stops if its rate of spread (or Byram intensity) is below a threshold, and goes through otherwise.
  2. A fuel layer that changes at a given time, so a road becomes a break from some hour on. It's simpler, but it needs the suppression timing as an input.

Is this something you already handle another way, or would you take a PR for 1?

Drafted with Claude Opus 5.5.

Activity

  1. filippi commented on Nov 3, 2026

    @filippi
    Contributor
  2. HugoFara commented on Nov 3, 2026

    @HugoFara
    CollaboratorAuthor

    Thanks. I tested momentum with a straight non-burnable strip on flat SH5 (Farsite, 5.3 m/s, 2 h), one strip across the wind for the head and one along it for the flank, in separate runs:

    relax minSpeed head crosses up to flank crosses up to free flank extent
    0.5 0 nothing, even 30 m nothing 0.8 km
    0.2 0 60 m 60 m 1.75 km
    0.2 0.05 30 m 30 m 1.7 km
    0.1 0 90 m 90 m 1.97 km
    0.1 0.05 60 m 60 m 1.8 km

    So the obstacle's width is the control, as you say, but head and flank cross at the same width. Lowering relax also makes the fire 2.5× wider (#253), so I can't tune it for this.

    Fire depth might separate them, but Farsite doesn't read it: only Balbi2015, BalbiUnsteady and FrontDepthDriven do. Would you make a break depend on depth inside one of those, or is there a way to use it with Farsite?

    Spotting would matter for this fire too (spot fires of 1–2 km). If you have a plan for it, I'd be glad to use Gonfaron as a test case.

    Drafted with Claude Opus 5.5, reviewed by a me before posting.

  3. added
    questionQuestions or investigations awaiting a diagnosis or decision
    on Nov 11, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    questionQuestions or investigations awaiting a diagnosis or decision

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions