Elmfire.jl

A Julia implementation of the ELMFIRE wildfire spread model. This package provides tools for simulating fire behavior including surface fire spread, crown fire, ember spotting, and more.

Features

• Rothermel Surface Fire Model: Industry-standard fire spread rate calculations
• Level-Set Solver: Accurate fire front propagation using narrow-band level-set methods
• Crown Fire: Coupled surface-crown fire modeling with canopy properties
• Spotting/Ember Transport: Lognormal ember distribution and transport physics
• Weather: Time-varying weather with spatial interpolation
• Geospatial I/O: Read/write GeoTIFF rasters, compute slope/aspect from DEMs
• Monte Carlo Ensembles: Probabilistic fire spread with burn probability maps
• Parallel Execution: Multi-threaded ensemble simulations
• WUI Models: Wildland-urban interface structure ignition
• Suppression Models: Containment line construction and resource allocation
• GPU Support: GPU-accelerated simulations via KernelAbstractions.jl

Installation

using Pkg
Pkg.add(url="https://github.com/RallypointOne/Elmfire.jl")

Quick Start

using Elmfire

# Create a 100x100 grid with 30-ft cells
state = FireState(100, 100, 30.0)

# Load standard fuel models (FBFM 1-13)
fuel_table = create_standard_fuel_table()

# Set weather conditions
weather = ConstantWeather(
    wind_speed_mph = 15.0,
    wind_direction = 270.0,  # Wind from the west
    M1 = 0.06,               # 1-hr dead fuel moisture (6%)
    M10 = 0.08,              # 10-hr dead fuel moisture
    M100 = 0.10,             # 100-hr dead fuel moisture
    MLH = 0.60,              # Live herbaceous moisture
    MLW = 0.90               # Live woody moisture
)

# Ignite the center of the grid
ignite!(state, 50, 50, 0.0)

# Run simulation for 30 minutes on fuel model 1 (short grass)
simulate_uniform!(state, 1, fuel_table, weather, 0.0, 0.0, 0.0, 30.0)

# Get results
burned_acres = get_burned_area_acres(state)

Crown Fire Simulation

# Enable crown fire with canopy properties
config = SimulationConfig{Float64}(
    enable_crown_fire = true,
    foliar_moisture = 100.0
)

simulate_full_uniform!(
    state, 1, fuel_table, weather, 0.0, 0.0, 0.0, 30.0;
    canopy_cbd = 0.15,  # Canopy bulk density (kg/m³)
    canopy_cbh = 3.0,   # Canopy base height (m)
    canopy_cc = 0.7,    # Canopy cover (70%)
    canopy_ch = 18.0,   # Canopy height (m)
    config = config
)

Time-Varying Weather

# Create weather grids for different times
grid1 = WeatherGrid{Float64}(weather_morning, 1, 1, 1e6)
grid2 = WeatherGrid{Float64}(weather_afternoon, 1, 1, 1e6)

# Create time series (weather changes at 60 minutes)
times = [0.0, 60.0]
weather_series = WeatherTimeSeries{Float64}([grid1, grid2], times)
weather_interp = WeatherInterpolator(weather_series, ncols, nrows, cellsize)

# Run with time-varying weather
simulate_full!(state, fuel_ids, fuel_table, weather_interp, slope, aspect, 0.0, 120.0)

Ensemble Simulations

# Configure perturbations for Monte Carlo runs
perturb_config = PerturbationConfig{Float64}(
    wind_speed_factor_range = (0.8, 1.2),
    wind_direction_std = 15.0,
    moisture_factor_range = (0.9, 1.1)
)

ensemble_config = EnsembleConfig{Float64}(
    n_simulations = 100,
    perturbation = perturb_config
)

# Run ensemble
result = run_ensemble!(
    ensemble_config, base_state, fuel_ids, fuel_table,
    weather, slope, aspect, 50, 50, 0.0, 60.0
)

# Get burn probability map
burn_prob = result.burn_probability

Geospatial Data

# Read landscape data from GeoTIFF files
landscape = read_landscape(
    fuel_path = "fuels.tif",
    dem_path = "elevation.tif"
)

# Write fire perimeter as GeoJSON
write_fire_perimeter(state, landscape.metadata, "perimeter.geojson")

# Write output raster
write_geotiff("burned.tif", state.burned, landscape.metadata)

Level Set PDE Solver

Fire front propagation is computed by solving the level set equation:

∂φ/∂t + F|∇φ| = 0

Where φ is the signed distance function (φ < 0 inside fire, φ > 0 outside) and F is the local fire spread rate. The numerical scheme uses:

• 2nd-order Runge-Kutta time integration
• Superbee flux limiter for gradient calculation (prevents oscillations)
• Narrow band method for efficiency (only computes near the fire front)
• CFL-adaptive timestep for numerical stability

Examples

See the examples/ directory for complete examples:

• basic_simulation.jl - Getting started with fire simulations
• spotting_example.jl - Ember transport and spot fire ignition
• animation.jl - Animated fire spread visualization
• marshall_fire/ - Case study with real-world data

Documentation

Tutorials are available in docs/tutorials/:

• Basic Simulation
• Weather Effects
• Terrain Effects
• Crown Fire
• Spotting / Ember Transport
• Ensemble Simulations
• WUI (Wildland-Urban Interface)
• Marshall Fire Case Study
• Suppression

References

• Rothermel, R.C. (1972). A mathematical model for predicting fire spread in wildland fuels.
• Anderson, H.E. (1983). Predicting wind-driven wild land fire size and shape.
• Scott, J.H. & Reinhardt, E.D. (2001). Assessing crown fire potential.

License

The Julia package code is licensed under the MIT License. The bundled ELMFIRE Fortran source code in the elmfire/ directory is licensed separately under the Eclipse Public License 2.0.