Imagine waking up early, eager to peep dazzling carpets of brilliant orange flowers at the Antelope Valley California Poppy Reserve. Instagram posts promised a spectacle.
You drive to the reserve north of Los Angeles, but the rolling hills aren’t alive with color.
Bummer. The bloom is over.
Thanks to AI, and a local scientist, such disappointment may soon be a thing of the past.
This year, Steve Klosterman, a biologist who works on natural climate solutions, launched a “wildflower forecast,” powered by a deep-learning model, satellite imagery and weather data.
In a sense, Klosterman, of Santa Monica, developed the tool to meet his own need.
Last spring, the Midwest transplant was hankering to see some wildflowers. He assumed there was some online resource that offered predictions or leveraged satellite images.
“Surely, there must be something,” he recalled thinking. “But there was nothing.”
There are tools. The state reserve operates a live cam trained on one swath of land. Theodore Payne, a California native plant nursery and education center, runs a wildflower hotline, where people can call in and hear weekly recorded reports on hot spots.
“These are all essential resources,” Klosterman said. “At the same time, they’re limited.”
Klosterman isn’t green when it comes to plants. His PhD, at Harvard, focused on the timing of new leaves on trees in the spring and color change in the fall.
For a class project, a team he was part of built a website that predicted those leaf changes in the Boston area. It was a hit.
To create the poppy bloom predictor, Klosterman turned to AI initially developed for medical imaging. He has harnessed it to instead analyze satellite images of the Antelope Valley.
The model scans 10-by-10-meter squares of land to determine whether poppies are present by their telltale orange color. (It also identifies tiny yellow flowers called goldfields.)
The model is trained on satellite images — which go back nine years — along with past weather data.
It then uses the current forecast, and recent flower status, to peer into the future.
If the mercury is going to hit 100 degrees and wind is picking up — and in previous years that led to withering flowers — that will guide the prediction.
Right now, the model can forecast five days out and is, as Klosterman puts it, “very much a work in progress.” It would be better, more powerful, if it had 100 years to learn from.
As more data are collected, it might someday be able to forecast a week or two out.
Right now, poppies are popping at the reserve in the western Mojave Desert.
It rained throughout the fall and into winter, and poppies need at least seven inches of rain to make a good showing, said Lori Wear, an interpreter at the reserve.
Snowfall in January seems to push them to another level, but that didn’t happen this season. So it’s a good bloom, but not extraordinary, she said.
Still, poppies — California’s state flower — blanket swaths of the protected land.
“It almost looks like Cheeto dust,” she said, “like somebody had Cheetos on their fingers and just smeared it on the landscape.”
Poppies here have typically peaked around mid-April, but variable weather in recent years has made it hard to predict, she said. Klosterman believes right now is likely the zenith.
Also blooming now: goldfields, purple grape soda lupine and owl’s clover. Wear described the latter, also purple, as looking like a “short owl with little eyes looking at you and a little beak.”
On Sunday, Klosterman experienced the blooms for himself, using his technology as a guide.
It offers predictions in two forms. The first is the amount of the valley — shown in a satellite image — covered in poppies and goldfields, expressed as a percentage. The other is an overlay of orange and yellow splotches on the land.
The map showed a fairly high concentration of poppies near a stretch of Highway 138. He went there and, lo and behold, vibrant flowers awaited him. He sent proof: a smiling selfie in front of a sea of blossoms.
Klosterman’s tool may help answer arguably more complex questions than poppy or no poppy, such as a more precise understanding of the conditions the flowers need to thrive.
Experts know rain is key, but it’s more complicated than that.
Heavy rain can supercharge invasive grasses, crowding out the blooms. Natives actually tend to do better after several years of drought, once invasives not adapted to the arid climate die out. That’s what led to an epic superbloom in 2017, Joan Dudney, an assistant professor of forest ecology at UC Santa Barbara, told The Times in 2024.
Klosterman wondered if the recent heatwave would desiccate them. But his model didn’t show that, and neither did his trip. So it’s possible other factors play a significant role in their persistence, such as length of day.
The model could also shed light on what could happen to the flowers as the climate warms. Will they migrate to the north? Will there be fewer blooms?
To game that out, Klosterman said you could invent and plug in a weather forecast with higher temperatures.
For now, Klosterman’s forecast is limited to the Antelope Valley. But if it expands to other areas, and other flower types, it could help people like Karina Silva.
Silva woke up at 5 a.m. last Wednesday to travel from her Las Vegas home to Death Valley National Park, hoping to beat the heat and the crowds to the superbloom.
But several hours later, she and her husband, David, were still trying to find it.
The hillside behind her was sprinkled with desert golds, but the display fell short of the riotous eruption of flowers posted on social media. The superbloom ended in early March, according to park officials.
“I was just thinking it was going to be this explosion of different colors,” Silva said by the side of the road overlooking Badwater Basin.
The post Don’t want to miss the bloom? This L.A. scientist created a poppy forecast appeared first on Los Angeles Times.
