Elmfire.jl

A Julia Implementation of the ELMFIRE Wildfire Simulation Model

Overview

Elmfire.jl is a pure Julia implementation of the ELMFIRE (Eulerian Level Set Model of FIRE spread) wildfire simulation model. It provides high-performance fire spread modeling with support for:

  • Rothermel Surface Fire Spread Model - Industry-standard fire behavior calculations
  • Level Set Method - Efficient fire front propagation using the Eulerian approach
  • Crown Fire Modeling - Passive and active crown fire transitions
  • Ember Transport (Spotting) - Stochastic ember generation and transport
  • Weather Interpolation - Spatially and temporally varying weather conditions
  • Monte Carlo Ensembles - Probabilistic fire forecasting with parallel execution
  • WUI (Wildland-Urban Interface) Models - Building ignition and urban fire spread
  • Suppression Models - Fire containment line construction and resource management
  • Geospatial I/O - GeoTIFF reading/writing for integration with GIS workflows

Features

Multi-Precision Support

Elmfire.jl supports both Float64 (default) and Float32 precision throughout the entire codebase, allowing you to trade off numerical precision for memory efficiency and potentially faster computation on certain hardware.

High Performance

  • Narrow Band Method - Only computes fire spread near the active fire front
  • Multi-threaded Ensemble Runs - Parallel execution of Monte Carlo simulations
  • Efficient Memory Layout - Column-major arrays optimized for Julia

Comprehensive Fire Behavior

  • 13 standard FBFM (Fire Behavior Fuel Models) included, with Scott & Burgan models loadable from CSV
  • Custom fuel model support
  • Dynamic fuel moisture effects
  • Slope and aspect terrain effects
  • Wind-driven elliptical spread patterns

Quick Start

using Elmfire

# Create a fire simulation state (100x100 grid, 30ft cells)
state = FireState{Float64}(100, 100, 30.0)

# Create standard fuel model table
fuel_table = create_standard_fuel_table(Float64)

# Set weather conditions
weather = ConstantWeather{Float64}(
    wind_speed_mph = 15.0,
    wind_direction = 270.0,  # From west
    M1 = 0.06,   # 6% 1-hr fuel moisture
    M10 = 0.08,
    M100 = 0.10,
    MLH = 0.60,
    MLW = 0.90
)

# Ignite at center
ignite!(state, 50, 50, 0.0)

# Run for 60 minutes on fuel model 1 (short grass), flat ground
simulate_uniform!(state, 1, fuel_table, weather, 0.0, 0.0, 0.0, 60.0)

# Get results
println("Burned area: ", round(get_burned_area_acres(state), digits = 1), " acres")
Burned area: 24.5 acres

state carries the outcome on the grid. Mapping the time of arrival shows the fire front’s progress: the wind blows from the west, so the fire runs east in the elongated ellipse the Rothermel and Richards models predict, with slow backing spread upwind.

using Plots

# Unburned cells are flagged with -1; mask them so they render as background
toa = [t < 0 ? NaN : t for t in state.time_of_arrival]

# Grid is indexed [x, y], so transpose to put east on the horizontal axis
xs = (1:state.ncols) .* state.cellsize
ys = (1:state.nrows) .* state.cellsize

heatmap(
    xs, ys, toa';
    aspect_ratio = 1,
    color = :inferno,
    xlabel = "Distance east (ft)",
    ylabel = "Distance north (ft)",
    title = "Time of arrival after 60 minutes",
    colorbar_title = "minutes",
    background_color_inside = :gray95,
)

Installation

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

Documentation Structure

Funding

This work was funded by the NSF CO-WY ASCEND Applied R&D Program.

License

Elmfire.jl is released under the MIT License.