Fire Alarm ML Prototyping

This repository contains code use to prototype ML models to estimate in situ air quality (PM2.5) from EPA sensors given estimated PM2.5 from GEOS-FP, as well as other meteorological and surface variables, along with VIIRS active wildfire pixels. It contains code to fetch these datasets, perform basic data processing and visualization, and model training/evaluation.

Contents

• config: model training and data fetching configuration files
• data: data used for model training or evaluation
• firealarm_ml: Python package containing scripts
• FireAlarm ML Prototype.ipynb: early prototype random forest model trained using data from the NOAA Hazard Mapping System.
• environment.yml: conda environment file