Benchmarking WUI Models

A research sandbox for benchmarking wildfire spread and impact models in the Wildland–Urban Interface (WUI). The repository bundles data-processing notebooks, model setups, and reproducible environments so you can compare different fire-behavior simulators side by side.

This is still a Work In Progress.

• Ready-to-run scripts and example configurations for Elmfire and WRF-Fire test cases.
• Notebook workflows for pulling ERA5 meteorological data and preparing inputs.
• Benchmarks multiple wildfire spread models focused on WUI scenarios, with a focus on the Maui, HI 2023 Wildfire.

Project structure

Data/                # ERA5 downloads, preprocessing notebooks, and helper scripts
Elmfire_Model/       # Elmfire build scripts, environment, and usage notes
WRF/em_fire/         # WRF-Fire test case files, namelists, and sample outputs

Quick start

1. Clone the repo

   git clone https://github.com/your-org/Benchmarking_WUI_Models.git
   cd Benchmarking_WUI_Models

2. Set up environments
   • For data preprocessing, use the Conda specs in Data/environment.yml.
   • For Elmfire, use Elmfire_Model/environment.yml and the load_elmfire.sh helper.
3. Run notebooks Launch JupyterLab from the Data folder to explore preprocessing workflows and ERA5 downloads.

Recommended workflow

1. Create the preprocessing environment:

   conda env create -f Data/environment.yml
   conda activate wui-data

2. Fetch meteorological inputs (ERA5) using Data/get_era5_data.py.
3. Open Data/ERA5_analysis.ipynb (or similar notebooks) to prepare forcing data for the models:

   cd Data
   ./start_jupyterlab.sh

Model setups

Elmfire

The Elmfire_Model folder contains build and run helpers for Elmfire.

1. Install Miniconda and source the helper script (ensure your scratch path is correct):

   cd Elmfire_Model
   source load_elmfire.sh

2. Build the model and tools:

   cd "$ELMFIRE_BASE_DIR/build/linux"
   ./make_gnu.sh

3. When running simulations, verify the scratch directory and Path_to_GDAL are correctly set in both data files and bash scripts.

WRF-Fire

The WRF/em_fire directory provides a WRF-Fire example case with namelists, input soundings, and sample outputs. To experiment:

1. Review WRF/em_fire/README.txt for WRF-specific background.
2. Use the provided namelist.fire and namelist.input variants to configure runs.
3. Run test scripts like run_testwrf.sh to generate output frames (e.g., wrfout_d01_*).

Initial Results

Generation of WRF meteorological inputs through super-resolving ERA5 data

Processing input layers for wildfire modelling

Initial run in Elmfire model