Balbi_2022_fixed_SFIRE model execution time using Anderson 2015 dataset
Workflow information
- Documentation page:
- Version: 1.0
- Date of record creation: 2024-12-03
- Date of upload to firebench: 2024-12-03
- Version/tag/commit firebench: 0.3.1a0
Configuration
- Rate of spread model: Rothermel using
firebench.ros_models.Balbi_2022_fixed_SFIRE
implementation. - Number of point Sobol: 2^10
Specific inputs
- The environmental variables chosen for this test are:
WIND_SPEED
from -15 m s-1 to 15 m s-1,SLOPE_ANGLE
from -45 deg to 45 deg,FUEL_MOISTURE_CONTENT
from 1% to 50%.
- Default values are used for optional inputs.
Hardware/software description
- Apple M2, macOS 14.7.1
- Python 3.10.8
Results
Fig.1 shows the execution time aggregated for all fuel classes (total) and for each fuel class. Minor discrepancies can be observed across fuel categories, mostly due to differences in the average number of iterations needed to converge. Overall, the performance is very similar for each fuel category, and a mean execution time of 19.11 \(\mu\)s over 106,496 samples.

Fig. 1 : Execution time boxplot for Balbi 2022 rate of spread model using Anderson13 fuel model. Fliers points not shown on the figure.
Data
- path to data:
- 01_generate_data.py:
f239568a63427ebb83fb00cfa955088e2b1d35abdb063794b08303f4a7cdcef8
- 02_plot_data.py:
3252272615d568451b75cc700b6afd6b3b939904fe370cb3f4dd6fdd8129caec
- 03_create_record.py:
cf7544abadcf6ccd0497e11a87e674d594c1500fe3c212e76177650372c6ba0f
- firebench.log:
d5104db25bf477348f8da4dd1f520a9c48a6ecd329c274478e7d45f56db9c683
- output_data.h5:
75673400b6ae742a9b5e11ddd7340e68e216f872057c638f899d9cd0afc05e1f
- efficiency_box.png:
6cdbb667bfb4a345c5f7597104d9db17af1432e9b061c404a8496c159f8a8552