Four days before the flames leapt across the ridge, Rohan Shivakumar, then a junior at Viewpoint School in Calabasas, California, noticed something curious about the trees outside. They were “sweating.” Not perspiring in the way humans do but exuding tiny droplets—clusters of moisture along leaf edges and twig joints—even in dry weather.
“It wasn’t rain, and it wasn’t dew,” he recalls. “Plants release water in response to an approaching fire threat. This release is detectable by instruments from space, and if we understand those trends multiple days before a fire hits a specific patch of vegetation, we can help predict fire spread before it happens. That has the potential to save lives, homes, and communities.”
When the wildfire finally descended four days later—its arrival sudden, its roar like a freight train—Rohan realized he had been watching more than just a fire. He’d been witnessing a warning.
Rohan’s fascination with wildfire patterns began long before this discovery. “In 2018, when I was ten years old, the Woolsey Fire ravaged my neighborhood,” he says. “While I was lucky to escape with my house intact, many of my friends and even my family members were not as lucky. I asked my parents questions about fire spread. Their answers sparked my interest to learn how to help communities, and in high school I decided to pursue my passion for helping people further.”
Growing up in Southern California, Rohan has always been surrounded by both the beauty and the danger of nature. “It’s been both a blessing and a curse to live in the hills,” he explains. “You have incredible scenery, but also the constant threat of wildfire. And living close to JPL and Caltech—both at the forefront of remote sensing—definitely influenced my interest in how technology can help.”
Once Rohan suspected the plant “sweating” was tied to wildfire risk, he set out to prove it. “There were two big steps to the process,” he explains. “First, get fire perimeters. Then, track evaporative stress over those perimeters.”
Using ECOSTRESS’s Land Surface Temperature maps, he identified the hottest areas—where fires were burning most intensely. To measure plant stress, he turned to satellites like Landsat, Sentinel-2, and GOES, accessing data through OpenET’s models. “That allowed us to see how stressed plants were, in terms of water release, and analyze how that changed as a fire approached.”
Working with Dr. Joshua B. Fisher, an environmental science expert he met during UCLA’s COSMOS program, Rohan honed the rigor of his research. “I feel incredibly lucky to have had Dr. Fisher to mentor me,” Rohan says. “He helped me design my procedures, gave feedback on my figures, and pushed me to make my analysis more robust. That kind of guidance is invaluable.”
One of the most profound takeaways for Rohan has been a shift in how he views plants themselves. “Through my research, I’ve learned to think of trees not just as objects but as teachers,” he says. “They have so much to tell us, even beyond wildfire applications. Who knows—they might be able to help us solve other big problems like drought or climate change someday.”
That philosophy drives his work beyond the lab. He hopes his findings can be integrated into real-time monitoring systems for firefighting agencies. “My mentor has worked with Urban Sky, a startup that launches high-altitude balloons to monitor fires. We’re exploring how to integrate our findings into their systems so firefighters can flag at-risk areas and allocate resources more effectively.”
Rohan admits the path wasn’t easy. “One of the biggest challenges was navigating missing or low-quality data,” he explains. “Much of the ECOSTRESS imagery I needed was incomplete or missing for certain days. I had to sort through hundreds of files to find usable data for each fire. It was tedious, but it taught me persistence.”
He’s also experienced the emotional weight of working on an issue so close to home. His goal, he says, is simple: “I hope my work can help firefighters make better evacuation orders and more effectively allocate resources during a fire.”
Now, as he prepares for college, Rohan is already working on the next stage of his research—a machine learning app that could predict wildfire spread in real time. “It’s based on artificial neural networks,” he explains. “The idea is to help firefighters understand where to allocate resources or order evacuations more effectively during active fires.”
His peers and mentors have responded with enthusiasm. “I’ve contacted researchers and professors around the country for feedback, and they’ve shown interest,” Rohan says. “We’re working to get published in a peer-reviewed journal.”
In college, he plans to major in environmental science or engineering to expand his work on wildfires, drought, and climate resilience.
What Rohan wishes more people understood is that nature is constantly signaling. “Originally, like many people, I thought fires were completely random and determined by environmental factors like lightning,” he says. “While that’s partly true, it’s astounding how much we can learn through technology. We can use remote sensing to understand fire spread and even integrate machine learning to predict it.”
He believes the more we listen, the more prepared we can be. “Plants don’t ask for attention,” he says, “but if we’re quiet enough to listen, they’ll tell us what’s coming.”
For Wildfire Prevention: By identifying early physiological stress in plants, Rohan’s method adds a new layer to traditional risk models, potentially buying precious time for action.
For Youth Innovation: His story is proof that driven students can push science forward in ways that impact real-world safety.
For Human-Nature Dialogue: He’s not just decoding data—he’s translating a biological language into human terms we can act on.
And while the future of GroveGuard and his machine learning models is still unfolding, one thing is clear: in the conversation about wildfire preparedness, Rohan Shivakumar has given the plants a voice.
