How hrishir746/wildfire-prediction works
This is a research-oriented wildfire-spread prediction system: environmental inputs feed a learned pixel-level predictor, with a Rothermel-inspired simulator as a deterministic baseline; predictions can be evaluated against historical fires and explained through a read-only agent. The README also documents training dashboards and historical-data workflows. The tree does not include the README-referenced src/data/real_data.py, so data preparation is represented by the available fetch script. Several script-level links reflect README statements or SOURCE INDEX references rather than inspected implementations.
Main entry points: unet.py U-Net Predictor · evaluate_real_fires.py Historical Fire Evaluation · train_extended_gpu.py Model Training · rothermel_simulator.py Rothermel Simulator
Components 18
Environmental and Fire Data
- Fetch Real Inputs
scripts/fetch_real_data.py - Environmental Inputs
- Historical Fire Data
scripts/fetch_mtbs_data.py
Simulation and Prediction
- Rothermel Simulator
src/simulation/rothermel_simulator.py - GPU Simulator
src/simulation/rothermel_simulator_gpu.py - U-Net Predictor
src/model/unet.py - FU-Net Predictor
src/model/fu_unet.py
Training and Monitoring
- Model Training
scripts/train_extended_gpu.py - FU-Net Training
scripts/train_fu_unet.py - Training Dashboard
scripts/training_dashboard.py
Evaluation and Interpretation
- Prediction Metrics
src/evaluation/metrics.py - Historical Fire Evaluation
scripts/evaluate_real_fires.py - Model Validation
scripts/wildfire_validation.py - Agent Interface
scripts/agent_app.py - Prediction Explainer
src/agent/agent.py
Research Extensions
- RL Environment
src/simulation/wildfire_rl_env.py - Fire Progression Data
scripts/process_firms_progressions_temporal.py
Other components
- Researcher
Connections 24
- Researcher → Fetch Real Inputs: requests inputs
README.md - Fetch Real Inputs → Environmental Inputs: produces inputs
README.md - Environmental Inputs → U-Net Predictor: provides channels
README.md - Researcher → Model Training: starts training
README.md - Model Training → U-Net Predictor: uses model
scripts/train_extended_gpu.py - Model Training → Rothermel Simulator: generates targets
scripts/train_extended_gpu.py - Researcher → Rothermel Simulator: runs simulation
README.md - Researcher → Historical Fire Evaluation: evaluates event
README.md - Historical Fire Evaluation → U-Net Predictor: uses predictor
scripts/evaluate_real_fires.py - Historical Fire Evaluation → Rothermel Simulator: compares baseline
scripts/evaluate_real_fires.py - Historical Fire Evaluation → Prediction Metrics: computes metrics
- Researcher → Agent Interface: asks questions
README.md - Agent Interface → Prediction Explainer: launches agent
scripts/agent_app.py - Prediction Explainer → U-Net Predictor: loads predictor
src/agent/agent.py - Prediction Explainer → Prediction Metrics: compares metrics
src/agent/agent.py - Researcher → Training Dashboard: monitors training
README.md - Training Dashboard → U-Net Predictor: loads checkpoints
- Historical Fire Data → Historical Fire Evaluation: supports validation
- FU-Net Training → FU-Net Predictor: trains model
scripts/train_fu_unet.py - Fire Progression Data → Model Training: may supply examples
- RL Environment → Rothermel Simulator: calls simulation
src/simulation/wildfire_rl_env.py - Researcher → Model Validation: checks predictions
README.md - Model Validation → U-Net Predictor: uses model
scripts/wildfire_validation.py - Researcher → GPU Simulator: runs accelerated simulation