This dataset was generated particularly for the Physics Guided Machine Learning (PGML) research and
educational tasks. It is an ensemble of prescribed fire simulations generated by the QUIC-Fire coupled
fire-atmospheric modeling tool. Each simulation run is represented by a zarr file, containing the outputs created
by QUIC-Fire through the BurnPro3D web interface. To model a burn, users upload the polygon for their burn unit
and BurnPro3D uses a 3D fuels model for that location created by FastFuels and an ignition file with a
user-defined ignition pattern created in DripTorch. Those files are included here as well. In addition,
users define the environmental conditions they would like to model in terms of fuel moisture, wind direction and
wind speed. The relevant information about the versions used of BurnPro3D, QUIF-Fire, FastFuels and DripTorch are
included in the metadata, along with both presets defined by BP3D and user-defined model inputs.
Common Parameters Across All Simulation Runs:
sim_time: 600
fuel: {'xlen': 600, 'ylen': 600, 'density': 0.7, 'height': 1}
output: {'steps_fire': 1, 'steps_wind': 1, 'energy_atmos': True, 'fire_energy': True, 'fuels_moist': True}
topo: {'total_startup_iters': 0}
Varying Parameters:
wind_speed: 7 unique values
wind_direction: 11 unique values
surface_moisture: 3 unique values
ignition: 5 unique values
Total: 1155 simulation runs.
Ignition Files:
Ignite_Aerial.dat
Ignite_LongFireline_Inwards.dat
Ignite_LongFireline_Outwards.dat
Ignite_Strip_Northwards.dat
Ignite_Strip_Southwards.dat
Metadata JSON File: uniform-pgml-success.bp3d.json
List of Simulation Runs: uniform-pgml-success_list_simulation_runs.csv