How Watch Duty is using Gemini to get wildfire alerts out faster
Watch Duty leverages a Gemini-powered tool to convert noisy fire radio into text, spotting dispatches in under 30 seconds. This collaboration reduces error rates and helps volunteers verify incidents quickly, ensuring faster alerts reach communities during critical wildfire events.
Tags
Quick summary
Watch Duty leverages a Gemini-powered tool to convert noisy fire radio into text, spotting dispatches in under 30 seconds. This collaboration reduces error rates and helps volunteers verify incidents quickly, ensuring faster alerts reach communities during critical wildfire events.
How Watch Duty is using Gemini to get wildfire alerts out faster
Wildfires demand speed. In the critical minutes following ignition, the difference between safety and danger often hinges on how quickly information can be processed and disseminated. For Watch Duty, a nonprofit organization dedicated to wildfire alerts, the traditional method of monitoring fire radio has long been a bottleneck. Now, through a collaboration with Google, the organization is leveraging Gemini to transform noisy, analog audio into actionable intelligence, significantly reducing the time between a dispatch and a public alert.
The Challenge of Analog Fire Radio
The primary obstacle in wildfire detection remains the reliance on decades-old analog Very High Frequency (VHF) radio systems. These systems are prone to significant interference. Engine noise, wind, static, and rapid-fire tactical jargon create a chaotic audio environment where critical information is often buried.
Watch Duty reporters currently monitor up to 20 feeds manually to cut through this noise. This manual process is labor-intensive and prone to human error, especially when volunteers are fatigued or listening to multiple channels simultaneously. The challenge is acute because the first word of a new fire still travels over these analog channels. If a fire ignites at night or in an area with no one actively listening, the initial dispatch can go unnoticed until it is too late.
Turning Static into Text with Gemini
To address these challenges, Google engineers and researchers embedded full-time with Watch Duty for six months. Through a Google.org Fellowship, they developed a new Gemini-powered tool designed to automate the transcription of fire radio.
The system was trained using more than 20,000 hours of past radio recordings. This extensive dataset allowed the model to learn to wipe out heavy static and translate specialized firefighting terminology into clean, searchable text. By automating the deciphering of noisy audio, the tool removes the added strain on volunteers, allowing them to focus on verifying incidents and getting alerts out to the public.
The collaboration resulted in a tool capable of flagging new wildfire dispatches in under 30 seconds. This speed is crucial; it ensures that a new fire is not missed simply because no one was listening at that specific moment.
Measurable Improvements in Detection and Accuracy
The deployment of the Gemini-powered tool has yielded significant quantitative improvements in Watch Duty's operations. The system has successfully transcribed more than 200 hours of radio speech a day from approximately 400 active feeds. Furthermore, the tool has been designed and tested to handle more than 10,000 feeds at once, providing a scalable solution for future fire seasons.
Perhaps the most critical metric is the reduction in error rates. The tool reduced the word error rate on degraded VHF radio audio from 38% down to 22.34%. This improvement ensures that the transcribed text accurately reflects the dispatches, reducing the risk of misinterpretation during high-stress situations.
Beyond technical metrics, the tool has had a profound impact on the human element of the operation. As a result of the collaboration, volunteer reporter satisfaction with radio transcription tools increased from 35.3% to 92.3%. This surge in satisfaction indicates that the technology is not only effective but also sustainable for the volunteers who rely on it.
Real-World Application: The Railroad Fire
The efficacy of the new tool was tested during a live fire event in July 2026. The Railroad Fire, a fast-moving wildfire, ignited at night in Washington. In this scenario, the first 911 dispatch went out over local fire radio, buried in static.
The Gemini-powered transcription tool transcribed the dispatch straight from the noisy radio audio and flagged it in under 30 seconds. This head start gave Watch Duty's team the necessary time to verify the fire and push a Critical Alert. This alert type is designed to break through do-not-disturb modes on user devices, ensuring urgent messages are seen immediately.
This event marked the first time Watch Duty had ever sent a Critical Alert overnight. The successful deployment proved that faster detection is vital when communities are asleep and every minute counts. The tool helped residents evacuate before flames reached homes, demonstrating the tangible life-saving potential of the technology.
Scaling Infrastructure and Search Integration
The success of the transcription tool is supported by broader infrastructure improvements and partnerships. Watch Duty sends push alerts through Firebase Cloud Messaging. As its audience grew to more than 16 million people last year, Google Cloud raised its messaging limit from 600,000 to 1 million to help alerts arrive without delay.
Additionally, Google Search has integrated Watch Duty into its emergency response protocols. During major US fires such as the Palisades, Eaton, Hurst, and Auto fires, Google Search featured Watch Duty in SOS alerts alongside official emergency guidance. This integration ensures that users searching for help or information during a crisis can find verified alerts quickly.
Funding and Broader Industry Partnership
The development of this tool is part of a deep, multi-year collaboration spanning Google's philanthropic, product, and research teams. In 2024, Google.org provided Watch Duty with $2 million in funding and $300,000 in Google Cloud credits. These resources helped compile satellite, camera, and radio data to validate new fires and support the development of the AI models.
As part of the AI Collaborative: Wildfires, Watch Duty is turning models and data into tools that agencies, communities, and responders trust and use. The goal is to scale this life-saving mission across all 50 states.
Conclusion
The partnership between Watch Duty and Google demonstrates the practical application of generative AI in public safety. By converting noisy analog radio into clear, real-time text, Gemini has enabled Watch Duty to detect wildfires faster and with greater accuracy. The reduction in word error rates, the increase in volunteer satisfaction, and the successful deployment during the Railroad Fire highlight the value of this collaboration. As the goal remains simple—getting people the warning they need while there is still time to act—these technological advancements offer a promising path forward for wildfire response across the country.
Source: https://blog.google/company-news/outreach-and-initiatives/google-org/watch-duty-gemini-tool



