PyroCast AI

It's important to take wildfire precautions in California! Our website allows users to input their address to predict the risk and severity of a potential fire using a custom trained ML model.

Comment

Inspiration

We were inspired by the devastating LA fires earlier this year to create a program to help communities better prepare for and respond to wildfires. The 2025 CAL FIRE data showed a significant spike in acres burned, despite a fairly average number of wildfires compared to previous years. This imbalance highlighted the urgency for smarter and faster tools.

What it does

Our website is a wildfire prediction and alert system that analyzes relevant environmental data such as temperature, wind speed, and historic outbreak areas to predict both fire risk and severity.

How we built it

We used sci-kit learn to train an AI model using RandomForestRegression, with predictors including location, temperature, day of year, and wind speed.

Challenges we ran into

It was extremely difficult to compile all the data needed for the AI model to be trained on. Specifically, geographic/temporal data with temperature and wind speed was hard to acquire.

Accomplishments that we're proud of

We are proud of having trained a machine learning model that can predict how much acres of fire would spread based on a particular location/time and the temperature/windspeed at that location and time.

What we learned

We learned that data is very valuable, and it is important to have reliable, cleaned data that is all pre-processed in order for the AI model to make predictions.

What's next for PyroCast AI

Next is going to be an evacuation plan based on the surrounding areas' fire risk and severity! We will inform users how to safely get away from a fire in the event that one occurs near them. We would also like to provide our AI model with more additional data to boost its accuracy.

Share this project:

Updates