Fire Intensity and spRead forecAst (FIRA)

Repository for a fire spread forecast model, generating hourly fire radiative power (FRP) prediction for air quality forecasting applications.

Required Python libraries

• Keras
• Matplotlib (for output plotting only)
• MetPy
• netCDF4
• NumPy
• pandas
• SciPy
• Tensorflow
• xarray
• XGBoost

Components

Trained machine learning models are available in model/fira_models.zip

Source Code (Python Script)	Script Description
fire_model	Main driver of FIRA
rave_preprocessor	Data preprocessor for initial FRP data from RAVE
fire_inputgen	Model input generator, including data re-gridding, fire frame selection and filtering, scaling, and normalization
fire_forecast	Driver for fire spread prediction model
fire_intensity	Driver for fire intensity prediction model
fire_mapgen	Fire mapper for grided FRP prediction generation
fire_evaluation	Visualization tool for grided FRP prediction

Namelist settings

Namelist Option	Option Description
Filename specification
frp_input	Filename of FRP input files
model_input	Filename of FIRA input files
model_output	Filename of FIRA output files
FRP source option
frp_source	Option of the source of initial FRP data (0: RAVE)
Simulation time settings
time_start	Simulation start time in UTC (YYYYMMDDHH00)
time_end	Simulation end time in UTC (YYYYMMDDHH00)
time_freq	Temporal frequency (in hour) of simulation
Simulation domain settings
lat_lower_left	Latitude of the lower left point of model domain (degree)
lat_upper_right	Latitude of the upper right point of model domain (degree)
lon_lower_left	Longitude of the lower left point of model domain (degree)
lon_upper_right	Longitude of the upper right point of model domain (degree)
Model options
opt_frpgen	Option for FRP preprocessor (0: off, 1: on)
opt_inputgen	Option for input generator (0: off, 1: on)
opt_forecast	Option for fire spread prediction model (0: off, 1: on)
opt_intensity	Option for fire intensity prediction model (0: off, 1: on)
opt_mapgen	Option for fire mapper (0: off, 1: on)
opt_evaluation	Option for result visualization (0: off, 1: on)
Manual/Fake fire settings
man_fire	Option for adding manual fire (0: off, 1: on)
man_lat	Latitude of manual fires, separated by , (degree)
man_lon	Longitude of manual fires, separated by , (degree)
man_lvl	Fire intensity (FRP) of manual fires, separated by , (MW)
Path for required input files
path_frp	Path for initial FRP files
path_elv	Path for terrain elevation files
path_ast	Path for surface type files
path_fh	Path for forest height files
path_vhi	Path for vegetation health index files
path_mete	Path for meteorological data files

Inputs

1. Filename format: [input name].[start time].f[simuluation time].nc input name and start time can be specified in namelist.

2. Input netCDF file components:

Variable Name	Variable Description
time	Time in UTC
INPUTLIST	List of input variables
input_noscale	Input variables per fire frames before scaling and normalization (units vary)
input	Input variables per fire frames after scaling and normalization (dimensionless)
frame_lat	Latitude of fire frames (degree)
frame_lon	Longitude of fire frames (degree)

3. List of input variables:

Variable Name	Variable Description	Data Source
Fire characteristics
frp	Fire radiative power (MW)	RAVE
Fuel/surface characteristics
elv	Terrain elevation (m)	GriddingMachine/Yamazaki et al. (2017)
fh	Forest height (m)	GEDI+Landsat (GLAD)
vhi	Vegetation health index (dimensionless)	VIIRS VHP
ast	Surface type (dimensionless)	VIIRS AST
Date/Time
doy	Day of year (dimensionless)	N/A
hour	Hour of the day	N/A
Meteorological conditions
t2m	2-meter temperature	HRRR/University of Utah
sh2	2-meter specific humidity	HRRR/University of Utah
prate	Precipitation rate (mm h-1)	HRRR/University of Utah
wd	10-meter wind direction (degree)	HRRR/University of Utah
ws	10-meter wind speed (m s-1)	HRRR/University of Utah

Outputs

1. Filename format: [output name].[start time].f[forecast hour].nc output name and start time can be specified in namelist.

2. Output netCDF file components:

Variable Name	Variable Description
time	Time of FRP prediction in UTC
INPUTLIST	List of input variables
frame_predic_ori	Original fire spread prediction per fire frames (dimensionless)
frame_predic_post	Post-processed fire spread prediction per fire frames (dimensionless)
frame_predic_frp	Fire intensity (FRP) prediction per fire frames (MW)
frame_lat	Latitude of fire frames (degree)
frame_lon	Longitude of fire frames (degree)
grid_predic	Gridded FRP prediction (MW)
grid_lat	Latitude of gridded FRP prediction (degree)
grid_lon	Longitude of gridded FRP prediction (degree)