Fire behaviour

Purpose

The fire-behaviour package replaces the heuristic fire-spread probability with a published surface-fire model. The first kernel is Rothermel (1972), with the standard corrections used by modern implementations. The inputs are explicit and metric; the outputs include rate of spread, reaction intensity, head fire intensity, flame length, and fuel consumption.

This page specifies the fuel, moisture, and environment contracts and the Rothermel kernel. It does not yet specify crown fire, the Canadian FBP fuel types, or landscape growth; those are later P156 tasks.

Fuel contract

A fuel model has five fuel classes: 1-hour, 10-hour, 100-hour, live herbaceous, and live woody. Each class carries:

  • oven-dry load in tonnes per hectare;

  • surface-to-volume ratio in square metres per cubic metre;

  • heat content in kilojoules per kilogram.

The model adds fuel bed depth in centimetres and the dead-fuel moisture of extinction in percent. The model type is S (static) or D (dynamic). A dynamic model transfers live herbaceous fuel to the 1-hour dead class as the live herbaceous moisture falls below 120 percent.

The 13 Anderson standard models are included as ANDERSON_13. The 16 Scott and Burgan moisture scenarios are included as MOISTURE_SCENARIOS; D2L2 is the default.

Moisture contract

FuelMoisture carries the five moisture percentages. Live herbaceous moisture must be zero or at least 30 percent, matching the Rothermel implementation’s validity range. A moisture above the dead-fuel moisture of extinction drives the dead reaction intensity to zero.

Environment contract

FireEnvironment carries:

  • midflame wind speed in kilometres per hour;

  • slope in percent;

  • optional wind direction, aspect, elevation, temperature, and relative humidity.

Only wind and slope enter the Rothermel kernel in this slice. The remaining fields are provenance for the FBP/weather work.

Rothermel kernel

surface_fire(fuel, moisture, environment) returns a RothermelResult with:

  • rate of spread in metres per minute;

  • dead, live, and total reaction intensity in kilowatts per square metre;

  • wind and slope coefficients;

  • heat source and heat sink;

  • packing ratio and relative packing ratio;

  • characteristic SAV and bulk density;

  • characteristic dead/live moisture and live moisture of extinction;

  • fuel consumed in the flaming zone;

  • Byram head fire intensity in kilowatts per metre;

  • Byram flame length in metres;

  • an estimated flaming residence time.

The calculation follows the model’s imperial formulation internally, as the reference implementations do, and converts the inputs and outputs to metric. The kernel is deterministic and has no hidden random draws.

Derived intensity

The kernel’s rate of spread and reaction intensity are the validated outputs. Head fire intensity and flame length are derived for the account and visual layers:

fuel_consumed_kg_m2 = total_load_kg_m2 × consumption_fraction
intensity_kw_m = heat_yield_kj_kg × fuel_consumed_kg_m2 × ROS_m_s
flame_length_m = 0.0775 × intensity_kw_m ^ 0.46

The default heat yield is 18,000 kJ/kg and the default consumption fraction is 0.8. These are explicit scene/account parameters, not part of the Rothermel validation.

Standard cases

The kernel is tested against reference values generated with the R Rothermel package (Vacchiano and Ascoli, 2015), which implements Rothermel (1972) with the Frandsen, Albini, and Andrews corrections. The package is a validation oracle only, not a runtime dependency. The standard 13 Anderson models at the D2L2 moisture scenario, 8 km/h wind, and level ground match within 0.01 in the reported units. The custom dynamic example and a 15-point A3 wind/slope grid also match within 0.01.

Example

python examples/fire_rothermel_demo.py

The demo prints a table for the 13 standard models at D2L2, 8 km/h wind, and level ground:

model  ROS m/min  reaction kW/m2  intensity kW/m  flame m
 A1      30.65         156.87         1221.2     2.04
 A3      39.12         550.01         6328.5     4.34
A10       2.97        1115.88         1920.7     2.51
A13       5.31        1847.20        16611.1     6.77

Programmatically:

from ws4.fire import ANDERSON_13, MOISTURE_SCENARIOS, FireEnvironment, surface_fire

result = surface_fire(
    ANDERSON_13["A3"],
    MOISTURE_SCENARIOS["D2L2"],
    FireEnvironment(wind_kmh=8.0, slope_percent=0.0),
)
print(result.rate_of_spread_m_min, result.head_fire_intensity_kw_m)

Limits

  • Only the Rothermel surface kernel is implemented. The Canadian FBP equations, crown fire, and landscape growth are later P156 tasks.

  • The standard cases validate the kernel, not a landscape. A standard-case pass does not prove a fire perimeter or an observed-fire match.

  • Head fire intensity and flame length are derived and are not part of the reference-case comparison.

  • The fuel models are the 13 Anderson models. The Scott and Burgan 40 and the FBP fuel types follow.

  • The kernel is deterministic; uncertainty belongs to the weather, moisture, and ignition ensemble.

Cross-references