Comparison with Other Wildfire Simulation Tools

Note

Note on naming: WRF-Fire and WRF-SFIRE refer to the same coupled fire–atmosphere system built on the WRF mesoscale model. The official package name is WRF-SFIRE; “WRF-Fire” is a common shorthand used in the community.

Warning

Disclaimer: The comparisons below are based on publicly available documentation and peer-reviewed literature as of 2025. Capabilities of third-party tools may have evolved or may differ across versions. Consult the official references linked in the Tool Documentation References section for authoritative and up-to-date information. This table is intended for high-level orientation only and should not be taken as an exhaustive or definitive characterisation of any tool.

Capability Summary

Tools are grouped into three columns to keep the table compact:

  • Group A — Operational fire behavior tools: FARSITE · FlamMap · BehavePlus

  • Group B — Physics-based LES/CFD tools: QUIC-Fire (QUIC-URB) · FIRETEC

  • Group C — Coupled fire–atmosphere NWP: WRF-Fire (WRF-SFIRE)

Capability

Wildfire-AMR (this solver)

Group A — FARSITE / FlamMap / BehavePlus

Group B — QUIC-Fire / FIRETEC

Group C — WRF-Fire (WRF-SFIRE)

Surface spread model

Rothermel (1972), Balbi (2009), and 5 others (Cheney–Gould, Cruz crown, FBP, Lautenberger, plus Viegas eruptive option)

Rothermel (1972) in all three

Semi-empirical QUIC; physics LES FIRETEC

Rothermel (1972)

Propagation method

Eulerian level-set (WENO5-Z / RK3); FARSITE Huygens ellipse; MTT

FARSITE: Huygens wavelet; FlamMap: MTT or Huygens; BehavePlus: point (no spatial propagation)

Eulerian CFD (QUIC-URB / HIGRAD-FIRETEC)

Eulerian level-set on WRF grid

Crown fire

Van Wagner (1977) + Rothermel (1991) + Cruz et al. (2005) + Scott–Reinhardt (2001) TI/CI

Van Wagner (1977); FARSITE/FlamMap active via Rc= 3.34 Rs; BehavePlus: Van Wagner point calc.

Physics-based combustion

Van Wagner (1977)

Wind adjustment

WAF (Andrews 2018); MEWS cap; 7 wind-terrain feedback models

FARSITE/FlamMap: WAF + MEWS (internal); BehavePlus: user-specified WAF

3-D mass-consistent (QUIC-URB) or LES

WRF-derived; WAF in coupling layer

Fuel models

FBFM13 + FBFM40; FBP grass/slash; Lautenberger; per-cell LCP

FARSITE/FlamMap: FBFM13 + FBFM40; BehavePlus: FBFM13 + FBFM40

Custom 3-D bulk density per cell

FBFM13

Fuel moisture

All size classes; FMD schedule; Nelson (2000) diurnal EMC; precipitation wetting; solar shading; per-cell .fms; spatial output in plotfiles

FARSITE/FlamMap: dead/live + FMD + conditioning; BehavePlus: dead/live per class

Bulk moisture per cell

Dead/live (prescribed)

Non-burnable masking

✓ Codes 91–99 / NB1–NB9 → ROS = 0

✓ (all three)

N/A (3-D grid)

Partial (fuel mask)

Flame diagnostics

Byram intensity + flame length; scorch height; tree mortality; TI/CI; NFDRS ERC

FARSITE: intensity + flame length; FlamMap: full outputs; BehavePlus: full outputs

Physics heat release

Intensity + flame length

Firebrand spotting

Albini (1983) 2-D trajectory + torching; stochastic distance model

FARSITE: Albini empirical; FlamMap: not standard; BehavePlus: Albini (point)

Select FIRETEC versions

Not included

Terrain & landscape

Per-cell elev./slope/aspect/fuel from LCP or XYZ terrain file; 2-D domain

All three: full 2-D LCP landscape; BehavePlus: user slope/aspect

Full 3-D terrain + canopy

Full 3-D WRF terrain

Weather input

Single .wtr; FMD schedule; multi-station IDW spatial interpolation

FARSITE: per-station .wtr; FlamMap: gridded or single station; BehavePlus: single point

Prescribed per-cell

WRF atmospheric profiles

Fire–atmosphere coupling

None (prescribed wind; heat-flux plume correction optional)

None in all three

One-way QUIC-URB; two-way LES FIRETEC

Two-way (fire ↔ WRF)

Barrier / suppression

Polyline firebreaks (CSV); aerial retardant (ROS + spotting suppressed)

FARSITE: dozer/hand lines + retardant; FlamMap/BehavePlus: not included

Not included

Not included

GPU acceleration

✓ AMReX CUDA / HIP / SYCL

✗ (all three are serial / CPU-only)

Partial (QUIC-URB select versions)

✗

MPI parallelism

✓ AMReX domain decomposition

✗ (all three serial)

✓ MPI/OpenMP

✓ WRF MPI

Embedded boundaries

✓ AMReX EB (buildings, fuel breaks)

✗ (all three)

✓ QUIC-URB / FIRETEC 3-D geometry

✗

Open source

✓ MIT licence

FARSITE/FlamMap: proprietary USFS binary; BehavePlus: open source

Research licence (QUIC-Fire); restricted (FIRETEC)

✓ WRF open source

Tool Documentation References

For authoritative and up-to-date information on each tool, refer to the official sources:

Key Differences from FARSITE

  • Eulerian level-set vs. explicit Huygens wavelets: Wildfire-AMR embeds the same Richards (1990) elliptical directional spread as FARSITE inside an Eulerian level-set (WENO5-Z/RK3). The fire perimeter is the zero contour of a signed-distance function; merging fronts and islands are handled automatically without explicit connectivity management.

  • Extended spread model library: In addition to Rothermel (1972), Wildfire-AMR includes Balbi (2009) physics-based, Cheney–Gould (1995) Australian grassland, Cruz et al. (2005) crown fire, Canadian FBP O1a/O1b/S1–S3 (Forestry Canada 1992), and Lautenberger (2013) physics-based. FARSITE ships only with Rothermel (1972).

  • Non-burnable cell masking: Fuel model codes 91–99 and NB1–NB9 (water, rock, urban, bare ground) are explicitly zeroed in the ROS kernel so fire cannot creep through sparse-fuel numerical noise into non-burnable areas.

  • Multiple weather stations: multi_wtr_file loads per-station .wtr files and produces spatially-varying wind and T/RH via IDW interpolation, matching FARSITE’s multi-station weather capability.

  • Retardant spotting suppression: retardant_file now suppresses both ROS and spotting probability inside active drop zones, consistent with FARSITE’s aerial retardant model.

  • Wind adjustments are optional: FARSITE applies WAF and MEWS internally. Wildfire-AMR exposes both via rothermel.use_waf and rothermel.use_wind_limit so users can match FARSITE behaviour or supply midflame-height wind directly.

  • GPU and MPI: AMReX CUDA/HIP/SYCL kernels and MPI domain decomposition. FARSITE is serial and CPU-only.

  • Embedded Boundary: AMReX EB allows buildings and fuel breaks on the Cartesian grid without remeshing. FARSITE has no EB support.

Key Differences from WRF-Fire (WRF-SFIRE)

  • No atmospheric coupling: Wildfire-AMR uses prescribed wind fields (constant, CSV, or WRF output); WRF-SFIRE fully couples fire with WRF, including fire-induced wind, heat flux, and smoke transport.

  • Wind adjustment: When driving with WRF output, enable rothermel.use_waf = 1 to convert NWP wind to midflame height. WRF-SFIRE handles this inside the coupled framework.

  • Richer fire behaviour models: WRF-SFIRE uses Rothermel (1972) only. Wildfire-AMR supports seven additional spread models and a richer crown fire pipeline (see above).

  • Simpler setup: Wildfire-AMR requires only CMake and AMReX; WRF-SFIRE requires a full WRF stack (NetCDF, MPI, WPS, WRF pre-processing).

  • GPU-native kernels: WRF-SFIRE is CPU-MPI; Wildfire-AMR uses AMReX GPU kernels throughout.

Key Differences from FlamMap

  • Time-dependent propagation: FlamMap computes static fire behaviour maps (no time stepping); Wildfire-AMR evolves the fire front dynamically via level-set, FARSITE Huygens, or MTT.

  • Crown fire depth: FlamMap provides the full Scott & Reinhardt (2001) crown fire assessment. Wildfire-AMR matches this with bisection-based TI/CI plus Van Wagner (1977) passive blending and Cruz et al. (2005) active crown ROS.

  • GPU / open source: FlamMap is a closed-source Windows binary; Wildfire-AMR is MIT-licensed and GPU-accelerated.