Jybbs/sfire-samples

Dataset Card for Thermur WRF-SFIRE Sample Data

This dataset provides a sample from the Moisseeva (2020) WRF-SFIRE LES Synthetic Wildfire Plume Dataset, specifically curated for development and testing of the Thermur project - an imitation learning system for coordinating drone swarms in wildfire monitoring scenarios.

Dataset Details

Dataset Description

This sample contains a representative slice of high-fidelity wildfire simulation data, capturing active fire spread with well-developed plume dynamics. The data represents minutes 3-4 of a 20-minute simulation, providing realistic thermal gradients and turbulence patterns essential for algorithm development.

• Curated by: Thermur Project Contributors
• Original data by: Nadya Moisseeva and Roland Stull (2020)
• Language(s) (NLP): Not applicable (numerical simulation data)
• License: CC-BY-4.0 (following original dataset license)

Dataset Sources

• Repository: Thermur Project
• Paper: Original dataset publication
• Full dataset: FRDR Repository

Uses

Direct Use

This dataset is intended for:

• Algorithm development: Testing thermal navigation algorithms for autonomous agents
• Visualization development: Creating 3D renderings of fire plumes and thermal fields
• Machine learning: Training models to predict safe flight paths through fire environments
• Education: Understanding wildfire dynamics and plume behavior

Out-of-Scope Use

This dataset should not be used for:

• Real-time wildfire response without validation on current conditions
• Direct deployment to production drone systems without extensive testing
• Climate modeling (single scenario only)
• Predictive modeling of actual wildfire events

Dataset Structure

The dataset contains a single NetCDF file (samples.tar.gz → data/samples/wrf_sample.nc) with the following structure:

Dimensions:
  Time: 4 (timesteps at 15-second intervals)
  bottom_top: 50 (vertical levels)
  south_north: 250 (y-direction, 10 km)
  west_east: 500 (x-direction, 20 km)
  
Variables:
  - U, V, W: 3D wind velocity components (m/s)
  - T: Perturbation potential temperature (K)
  - P, PB: Perturbation and base pressure (Pa)
  - QVAPOR: Water vapor mixing ratio (kg/kg)
  - GRNHFX: Ground heat flux from fire (W/m²)
  - FGRNHFX: Fine-mesh ground heat flux (W/m²)
  - tr17_1: PM2.5-like passive tracer (μg/kg)

Physical Scenario

Moderate-intensity wildfire:

• Wind: 5 m/s uniform horizontal (moderate breeze)
• Fuel: Southern rough (dense brushy understory)
• Fire intensity: ~10-20 MW/m fireline
• Domain: 10 km × 20 km × 3 km
• Resolution: 40 m horizontal, 4 m fire mesh

Dataset Creation

Curation Rationale

This sample was extracted to provide a manageable dataset (468 MB) that still captures the essential complexity of wildfire plume dynamics. The selected timeframe (minutes 3-4) represents established fire behavior with active spread and well-developed thermal structures.

Source Data

Data Collection and Processing

The data comes from Large Eddy Simulations (LES) using WRF-SFIRE, a coupled atmosphere-fire modeling system. The original simulation:

• Used real atmospheric profiles from North American soundings
• Implemented Anderson fuel models for realistic fire behavior
• Ran at 40m atmospheric resolution with 4m fire mesh refinement

Who are the source data producers?

Original simulations by Nadya Moisseeva and Roland Stull at the University of British Columbia, using WRF-SFIRE developed by NCAR and University of Colorado Denver.

Bias, Risks, and Limitations

• Single scenario: This sample represents only one of 147 available scenarios
• Idealized conditions: Simulations use simplified terrain (flat) and uniform fuels
• Temporal snapshot: Only 1 minute of a 20-minute simulation
• Model limitations: LES approximations may not capture all fine-scale turbulence

Recommendations

• Use this data for algorithm development and testing only
• Validate algorithms on multiple scenarios before deployment
• Consider environmental variability not captured in this single sample
• Combine with real-world data when available

Citation

BibTeX:

@dataset{moisseeva2020wrfsfire,
  author    = {Moisseeva, Nadya and Stull, Roland},
  title     = {WRF-SFIRE LES Synthetic Wildfire Plume Dataset},
  year      = {2020},
  publisher = {Federated Research Data Repository},
  doi       = {10.20383/102.0314},
  url       = {https://doi.org/10.20383/102.0314}
}

APA: Moisseeva, N., & Stull, R. (2020). WRF-SFIRE LES Synthetic Wildfire Plume Dataset. Federated Research Data Repository. https://doi.org/10.20383/102.0314

More Information

For detailed variable descriptions, safety thresholds, and processing examples, see the full README included in the tar.gz file. The Thermur project documentation provides extensive examples of using this data for drone swarm coordination algorithms.