As profiled by UF News, the innovative Gainesville-based company Fire Neural Network, or FNN, is using artificial intelligence to reduce the time it takes to identify a devastating wildfire from 24 hours to just 40 seconds. ECE Florida has deep connections to this work—underpinning FNN’s ingenious technology is a deep understanding of lightning, its behavior, and its electrical signatures, based on decades of work at two labs in the ECE Department. Both the Ionospheric Radio Lab, directed by Dr. Robert Moore, and the International Center for Lightning Research and Testing (ICLRT), directed by Distinguished Professor Vladimir Rakov, have contributed mightily to the framework which allow FNN’s devices to make sense of the sensor data coming from the field. While FNN leads the NSF grant powering the work, ECE Professor Yong-Kyu Yoon leads a Small Business Innovation Research (SBIR) project which is a sub-award to the NSF project. In yet another ECE connection, Dr. Rakov was chair of FNN founder Istvan Keresz’s Ph.D. committee.

The Background

The abstract of the NSF project, “Extremely-Low-Frequency Characterization of High-Risk-Lightning,” clearly lays out the rationale driving this work:

Over 70% of the area burned in the Western US is due to lightning-initiated wildfires, which on average are nine times larger than human-initiated fires. On a national level, lightning-initiated wildfires burned 5.5 million acres in 2020, making lightning the number one cause of wildfires in terms of area burned. Lightning ignitions were responsible for more area burned than all other ignition causes combined. However, less than 10% of lightning strikes are capable of starting a fire, thus it is essential to correctly pinpoint High-Risk-Lightning (HRL™) in real-time. Yet, the technology currently available to fire management agencies neither quickly nor accurately identifies lightning ignitions.

With the funding provided by NSF, Fire Neural Network aims to address this need by conducting fundamental research on extremely low frequency (ELF) emissions of long-continuing-current (LCC) lightning to prove the feasibility of a product that will allow the detection and location of likely fire ignition spots in 40 seconds. Such rapid and accurate prediction would greatly reduce firefighter response times, enable efficient and focused fire mitigation strategies, and thus diminish the number of uncontrollable fires.

The Secret Sauce

The Ionospheric Radio Lab has a long history of providing scientific-grade measurements of ELF radio signals associated with lightning, having specialized in this work for decades. In supporting FNN, the IRL will analyze the ground-wave and sky-wave signals emanated by the lightning detected by FNN and then provide FNN with an ELF calibration based on their experimental observations. The ten percent of lightning strikes that are capable of starting a fire have a very specific ELF signature, and IRL’s calibration information enables the detection of those dangerous strikes. The combination of the sensors deployed by FNN, the AI which pinpoints the lightning location, the calibrations provided by IRL, and the datasets provided by the ICLRT all work together to make the accurate prediction of wildfire possible.