ForeFire Wildfire Risk Modeling Exercise Submission

This repository contains a ForeFire-based reproducible submission for the Wildfire Risk Modeling Exercise. It uses the provided WildCommons synthetic Forest and Prairie scenario data, runs native ForeFire propagation models through Python bindings, exports GeoJSON perimeter outputs, and overlays those perimeters on scenario building polygons to estimate structures at risk.

The full submission narrative is in REPORT.md.

The workflow was generated and documented with an agent prompt in agent.md. Running that prompt with a capable coding agent gives the instructions needed to reproduce the report, generated GeoJSONs, benchmark files, and ForeFire execution notes.

Preview

Town of Forest, Farsite-like ForeFire Run

Town of Prairie, Farsite-like ForeFire Run

Key Files

• REPORT.md: final report with workflow, model architecture, outputs, performance, ForeFire installation notes, and coupled-atmosphere limitations.
• agent.md: reproduction prompt for generating the report and model outputs with an agent.
• wildcommons_farsite.py: main ForeFire/WildCommons runner.
• wildcommons_sim.py: single-model runner used by the benchmark.
• benchmark_forefire_performance.py: end-to-end performance benchmark.
• params.ff: ForeFire-style parameter snapshot for the submitted run configuration.
• submission/: clean Farsite-like GeoJSON submission files organized by Forest and Prairie scenario.
• export_submission_geojson.py: exporter/validator used to normalize the submission GeoJSON files.

Reproduction Summary

The workflow requires a Python environment where pyforefire imports successfully. ForeFire is native C++ software with Python bindings, so a standard hosted notebook image is not enough unless ForeFire has already been compiled and installed there.

python -m py_compile wildcommons_farsite.py wildcommons_sim.py benchmark_forefire_performance.py
python wildcommons_farsite.py --dataset Forest
python wildcommons_farsite.py --dataset Prairie
python benchmark_forefire_performance.py --datasets Forest,Prairie

The benchmark included in the report ran 10 model/dataset combinations in 376.41 seconds on the recorded local CPU environment.

Coupled-Atmosphere Scope

This submission does not run coupled ForeFire/Meso-NH simulations. The exercise data provide synoptic/tabular weather forcing, but not the 3D weather fields, vertical profiles, boundary conditions, or coupling files required for a coupled fire-atmosphere run.