Wildfire Damage Modeling

This repository contains our DATASCI 112 final project on neighborhood-scale wildfire damage modeling.

Project overview

We study whether pre-fire terrain, vegetation, and fuel context can help predict where structural damage is more likely after a wildfire. We aggregate structure-level damage inspection records to H3 hexagons, engineer terrain and LANDFIRE-based features, build visualizations, and compare classification models for hex-level damage prediction.

Our workflow is organized into three main notebooks:

• data_extraction_and_feature_engineering.ipynb
  Builds the hex-level dataset by aggregating damage inspections and attaching terrain, vegetation, and fuel features.

• data_visualizations.ipynb
  Produces map-based visualizations and exported figures from the cleaned wildfire context data.

• wildfire_damage_model_selection.ipynb
  Compares classification pipelines and selects a final model for wildfire damage prediction.

Authors

Markus Hoehn and Leonard Collomb

Poster