SCC-IRG: Ember Intelligence: Building Wildfire-Resistant Communities using Unmanned Aircraft Systems and Distributed Deep Learning

Uncontrolled wildfires are among the most destructive natural hazards. They threaten lives, infrastructure, and ecosystems. A critical but understudied driver of their spread is the ember, a burning particle that detaches from the main fire; travels with the wind; and ignites spot fires far ahead of the main fire line. Because ember behavior is difficult to observe during actual fires, it remains poorly understood and underrepresented in operational fire models. This Smart and Connected Communities Integrative Research Grant (SCC-IRG) project aims to develop a technological solution called Ember Intelligence. This is a new class of intelligent tools that detect, track, and forecast ember movement using drone-mounted infrared sensors, edge computing, and artificial intelligence. The broader objective of the work is to integrate fire science, unmanned aircraft systems, and machine learning into a system that supports real-time situational awareness, improved fire prediction, and more effective decision-making for communities at risk. By working closely with the US Department of Agriculture's Forest Service and public safety stakeholders, the project also engages with local communities near prescribed burn sites, such as Richfield, Utah, to align technology with operational needs and strengthen resilience. Educational outreach will involve learners from K–12 through postdoctoral levels, promoting engagement in STEM and wildfire response.

To realize this vision, the project develops a distributed system for real-time data collection and spatiotemporal modeling of ember transport dynamics. Thermal imaging data collected by sensor laden drones will be used to detect embers, fire fronts, and terrain conditions, with a goal of detecting and estimating ember velocity vectors using deep learning methods. These signals will inform self-supervised and federated learning algorithms that model ember motion and identify likely ignition zones under varying wind and topographic conditions. To overcome the connectivity limitations typical in wildfire zones, the research creates decentralized learning and model-distributed inference schemes that enable local adaptation and collaborative intelligence across airborne and ground platforms. The system will be validated using high-intensity prescribed burns conducted through the Fire and Smoke Model Evaluation Experiment proscribed burn which provides rare, ground-truth data that can be used to calibrate and test the models under real operational conditions. Beyond wildfires, the methods developed are expected to have broader applicability in environmental monitoring, autonomous sensing systems, and emergency response under extreme conditions.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.