Realtime Wildfire Mapping Across America
FireGrid is a personal project I’ve been building in my spare time over the past year. It’s a wildfire risk intelligence platform that combines real-time data, sensor inputs, and machine learning to help communities anticipate and respond to fire threats. It’s still in its early stages, but it’s the project I return to whenever I have time to build something that matters to me.
Project Summary
The platform visualizes both active and historical wildfires across the United States. It overlays that information with live weather data and topography to generate local risk maps. Alongside the web platform, I’m developing a small network of environmental sensors that detect early signs of ignition through changes in air composition, temperature, humidity, and sound.
Origins and Motivation
The idea for FireGrid came from my own experience growing up in Northern California, where wildfires have become a constant part of life. In 2021, my family helped my grandfather evacuate his cabin in the Santa Cruz Mountains as flames moved over the ridge behind us. There was no alert or warning, just a rush to leave. That moment stayed with me, and FireGrid became my way of trying to design a system that could make those minutes count for more.
What began as a small experiment with public fire data grew into a platform that integrates software, sensors, and modeling. Over time, it has become a way to explore how technology might give communities earlier, more reliable warnings.
Technical Overview
FireGrid’s current version includes several parts that I continue to develop and refine:
- Interactive Mapping Platform: Built with React, Mapbox, and PostgreSQL, it visualizes active and past fires, weather patterns, and hazard layers in real time.
- Risk Modeling: I’m training a model in PyTorch to detect the earliest indicators of ignition from the sensor network. It analyzes the acoustic signature of fires, such as the crackling of burning vegetation, along with chemical data from gas sensors that track changes in carbon monoxide and volatile compounds. The system is still experimental, but early results suggest it can distinguish small ignition events from normal background signals.
- Hardware Prototyping: Each FireGrid Sentry sensor uses an ESP32 microcontroller equipped with gas, temperature, humidity, and acoustic sensors. I’m testing them in controlled conditions and refining data transmission and power efficiency.
- Next Steps: I plan to deploy a small pilot network in the Santa Cruz Mountains to integrate live sensor data with the mapping platform and evaluate detection accuracy.
Recognition and Reflection
Earlier this year, FireGrid was selected as one of 35 finalist projects for the 776 Fellowship, out of more than 1,600 applicants. Although I couldn’t accept the offer because it required pausing my studies, the experience confirmed that the idea resonated beyond the personal level.
FireGrid is still a work in progress, built between classes and research, and developed mostly from curiosity and persistence. The system is far from complete, but each version moves it closer to what I set out to build: a tool that can give people more time to act when the next fire begins.