America Is Burning, But We're Still Flying Blind
Right now, the United States is grappling with a surge of wildfires, particularly across the Western states. Thousands of acres are being consumed in a matter of hours. Hundreds of structures are at risk, entire communities have been evacuated, and fire crews are stretched thin across rugged, remote terrain.
The dangerous reality of hotter summers, drier fuels, and more erratic weather is driving wildfires fueled by just the right explosive conditions. Trees bent, dust swirled, and with it comes the embers—tiny sparks lifted high into the sky and flung miles ahead of the main fire. One lands on a rooftop. Another in a patch of dry brush behind a home. Within minutes, those sparks become flames, and the flames become a wall of fire.
And it’s a story that’s repeating itself, again and again. Yet another summer season where we are not preventing fire from becoming catastrophic.
Despite the proliferation of satellites, drones, and aircraft, we were still making decisions with delayed, fragmented, or incomplete data. Some of the most widely used sources for fire detection have severe limitations in temporal resolution, passing over the same area only a couple times per day, and often missing the most critical moments—when the fire is growing rapidly, shifting directions, or threatening new communities. By the time a hot spot is detected, it may already be outdated. And when decisions are being made in minutes—not hours—a 9-to-12-hour delay can be catastrophic.
Current fire maps tell us where the fire was, not where it’s going. They are typically handed off in static formats and not integrated into predictive models that simulate fire spread based on wind, terrain, fuel conditions, and ember travel. Without modeling and temporal forecasting, responders are left without answers to the most critical questions:
Which lives and homes are most at risk in the next 3 hours?
Where should we place resources next?
How much of our resources do we need?
Accurate Predictive Modeling with WindTL
To solve this problem we built WindTL, a wildfire intelligence platform designed to forecast the very thing that makes these fires so unpredictable: embers. WindTL models how embers travel through the air—identifying not just where the fire is now, but where it’s most likely to go next, helping responders predict and prevent spot fires before they ignite. With WindTL, fire crews, emergency managers, and utility companies can log into the web, and easily and quickly gain a forward-looking view—allowing them to pre-position resources, issue early warnings, and protect critical and vulnerable infrastructure in the path of invisible, airborne threats. Over the past few days, we have put WindTL to work over several fires in Utah and Colorado.
Monroe Canyon Fire
In the morning of July 13, the hot, dry, and high wind conditions were adding up to ignite the Monroe Canyon Fire. As the first satellite hotspot was detected, the WindTL tool started to produce warnings: the high risk pre-fire conditions (left) coupled with the increasing winds (right) were fueling a rapidly spreading fire.
While fire propagation through the ground (shown in blue), offered by many tools out there, indicated slow fire growth, the WindTL tool provided an insight no other is capable of: very quickly the fire will be growing and spreading south, as seen on the yellow, orange, and red dots showing the potential for ember spread:
As the satellites came the next day, the prediction by WindTL materialized. The fire had quickly spread to the south:
Unfortunately, the lack of real-time data precluded WindTL from hour-by-hour validations and updated predictions that are only based on ground truth inputs when a satellite passes over the fire. In other words, this kind of critical piece of intelligence is only available every 12 hours.
Deer Creek Fire
When the Deer Creek Fire ignited on July 10 near La Sal, UT, conditions were primed for rapid spread (below, with a general moderate risk in yellow and areas of high risk for fire growth and ember cast in orange).
Winds were blowing steadily toward the northeast, and within minutes of ignition, WindTL began generating predictive models of ember cast. The first forecast revealed a risk of embers spotting well north of the fire, where dry fuels and rugged terrain could allow small ignitions to take hold undetected.
But WindTL saw more than that. Three hours later, it identified not only the northeastern ember threat, but also a southern “finger” of fire growth potential (left). When analysts compared this to WindTL’s pre-fire risk map (right), it became clear: that exact southern corridor had already been flagged as a high-risk escape route, highlighted in orange before the fire ever began.
Then came the test. With no new aerial mapping or perimeter updates for the next 12 hours, WindTL continued to run forecasts based only on its previous outputs—extrapolating from earlier conditions rather than ground-truth fire location. When the next satellite pass finally arrived, it confirmed everything WindTL had warned: the fire had escaped south, and embers had ignited spot fires ahead of the front in the north.
Turner Gulch Fire
When lightning struck the high desert southwest of Gateway, Colorado, on July 10, few could have predicted that it would ignite what would become the largest wildfire currently burning in the state. But behind the scenes, a quiet shift was taking place in how wildfires are tracked and forecasted.
Initially, the Turner Gulch Fire grew explosively, fueled by high winds, steep terrain, and dry fuels—scorching over 14,000 acres within days. The pre-fire risk tools from WindTL depicted an increasing risk due to evolving fire-prone conditions (observed in the high density of orange dots over the area).
Within hours of ignition, WindTL began generating forecasts, ingesting the latest atmospheric data, terrain features, and vegetation intel. The very first WindTL run, just after ignition on July 11 at 7 AM, identified a clear threat: intense westerly winds, clocking up to 50 mph, were driving embers eastward across the valley, by the yellow (25 to 50 mph winds) and green arrows (12 to 25 mph winds).
By 11 AM, the model showed those embers traveling further and across the valley, potentially creating spot fires, while the fire front was being pushed uphill and west over the ridge.
When the next satellite ground truth arrived later that day (red squares), it confirmed what WindTL had already shown. The fire had expanded in both directions, just as predicted.
By July 12, conditions began to shift. WindTL detected a wind reversal, with breezes beginning to push west ahead of the fire—complicating containment efforts.
Yet again, WindTL’s forecasts held: the satellite data at 7 PM validated the shift in fire behavior.
Over the following days, the system continued to monitor the fire’s trajectory. WindTL’s July 16 prediction showed a decrease in fire risk and negligible new ember activity.
But perhaps the most striking part of this story is what could have happened. For 12 hours at a time, WindTL had to predict the fire’s future without any new ground truth, relying solely on its earlier outputs. Yet, WindTL consistently and accurately predicted the direction, speed, and behavior of the Turner Gulch Fire. Now, imagine if the fire perimeter had been updated every hour—streamed live from aircraft and UAVs. This isn't a technologist's dream—it’s reality, as we demonstrated in a live demo integrating WindTL's fire modeling with real-time aerial imagery.
Real-Time Mapping and Forecasting with Trident Sensing
The predictions shown in these fires weren’t guesswork. This was physics-based, terrain-aware, real-time forecasting. And it raises the question:
What if WindTL had access to live fire perimeter updates every hour instead of every 12?
After several days of making predictions off of other predictions and only gaining intelligence every 12 hours, the integration between WindTL and partner Trident Sensing became critical. Trident Sensing deployed its advanced aerial sensing platform to fly over the Monroe Canyon, Deer Creek, and Turner Gulch Fires, capturing high-resolution infrared and multispectral imagery in near real-time. While traditional mapping efforts often take hours to process and distribute, Trident’s flights delivered live perimeter updates within minutes, feeding directly into WindTL’s modeling engine.
Fortunately, the efforts of the 475 firefighting personnel deployed came to fruition and the rain finally came, resulting in significant decrease of fire behavior of the Monroe Canyon, Deer Creek, and Turner Gulch Fires. As Trident Sensing flew over the Turner Gulch Fire on July 17 at approximately 1 PM EDT, the decreased fire behavior was evident from their real-time fire maps (see the dispersed hotspots over the Turner Gulch Fire below).
However, WindTL received this data directly from the airplane and ingested that into its modeling engine. This partnership transformed static snapshots into a dynamic intelligence stream—enabling WindTL to generate near-real-time forecasts of fire spread and ember propagation just as the aircraft was mapping the fires, providing true situational awareness and anticipated fire movement before it happened, making proactive decisions in real time instead of reacting hours later. First, it was able to validate its prediction for the previous night, where it showcased rapidly decreasing risk, and negligible ember and fire growth. But most importantly, it was able to run the fire models using data from the aircraft in near real-time, as Trident Sensing flew above the fire.
Imagine a world where every asset deployed to a wildfire—not just a few—became part of a real-time intelligence network. Every drone, helicopter, and mapping aircraft wouldn’t just collect data for post-mission reports, mitigation efforts from ground and air assets —they stream it live into WindTL. As the fire burns, WindTL continuously ingests perimeter updates, thermal imagery, suppression efforts and atmospheric readings, instantly updating forecasts for surface fire spread, ember trajectories, and areas at risk. Firefighters on the ground would no longer wait hours for a static map—they’d have a living, breathing fire model in their hands, showing them where to go, what to protect, and where the fire is headed next. With every aerial asset connected, WindTL turns fragmented efforts into a unified, predictive response—transforming wildfire management from reactive to truly proactive.
🔥 To Recap
The Turner Gulch Fire isn't just a test of firefighting endurance. It is a proving ground for a new generation of wildfire intelligence. One where planes don’t just observe, they inform. One where models don’t lag behind the fire—they run ahead of it. And one where WindTL and Trident Sensing turned a chaotic, wind-driven blaze into something we could finally see coming.
The partnership between WindTL and Trident Sensing marks a breakthrough moment in wildfire management—a shift from delayed reaction to real-time foresight. We may be bold, but we can share today that this has been one of the first times fire modeling has been performed off of live data from a mapping airplane.
By combining Trident’s rapid aerial mapping capabilities with WindTL’s next-generation predictive models, fire data that once took hours to collect, process, and distribute is now transformed into actionable intelligence within minutes. This collaboration enables incident commanders to see not just where the fire is, but where it’s going—empowering them to make faster, smarter decisions about evacuations, resource deployment, and containment strategies. In an era where fires are faster, hotter, and more destructive, the ability to close the loop between sensing and prediction is nothing short of transformational. Together, WindTL and Trident Sensing are redefining what’s possible on the front lines of wildfire response.