Predicting Utah snow can be tricky.
A person can’t physically measure the fresh powder falling on every frozen mountainside. Forecasters instead rely on computer algorithms to estimate how many inches a storm will bring and how wet the snow is.
But those forecast models are only as good as the real-world data they’re based on.
That’s why Peter Veals and other University of Utah scientists have created a new algorithm using data from more than a dozen snow sites in Utah and surrounding states. Early research published in 2025 indicates it could pay dividends.
“We found that we had a dramatic improvement in skill with this new approach relative to the old algorithm,” Veals said. “So, this will definitely improve snowfall forecasts in the West.”
Even a small upgrade in accuracy would be welcome news for anyone from skiers waiting for powder days to kids hoping for snow days to water managers tracking the state’s precious H20 supply.
There are a number of challenges to forecasting snow, said Veals, who is a research assistant professor of atmospheric sciences.
The big one is estimating how much water the snow holds. To do that, you need to know the right snow-to-liquid ratio — something Veals said is notoriously tricky to pinpoint.
On top of that, traditional forecast models were often trained on snowstorms in the eastern U.S., he said. That can be a problem because snow density in the West is more variable — from exceptionally dense wet snow in the Pacific Northwest to dry fluffy powder in Utah. The Eastern-trained models may not account for how mountains impact air movement during storms and where the snow is produced within the storm.
“If you don't have good data, you're not going to get good predictions,” Veals said. “That's why this was a relatively low-hanging piece of fruit to improve, because no one had really taken this approach of getting really high-quality data and then making a predictive model with that.”
Veal’s team used data collected from 14 snow sites in the mountains of Utah and several other Western states beginning in 2018. The Utah sites were located in Little Cottonwood Canyon, Big Cottonwood Canyon and Provo Canyon. Ski resort snow safety professionals and public safety teams manually collected measurements once or twice a day using a tube on top of a board. They then weighed the snow to determine its water content.
Veals’ team plugged the numbers from the Western mountains into the system they built, a form of artificial intelligence that learns and recognizes patterns.
“You tell this machine learning model to make a predictive algorithm that says, ‘Hey, for a given temperature, wind speed and humidity, what is the [snow-to-liquid ratio] going to be?’”
He expects the algorithm to be integrated into National Weather Service forecasts in the coming year. Shortly thereafter, it would start to inform the updates in Utahns’ weather apps.
It could also help keep people safe on the roads.
“We use weather forecasting for pretty much every decision that we end up doing,” said Steven Clark, avalanche safety program manager with the Utah Department of Transportation, which also contributed snow data for Veal’s model. “Arguably, our biggest decision-making tool is weather models.”
That’s why even a little bit more accuracy over previous models could be really valuable.
“It's not just for skiers and snowboarders,” Clark said. “This is really anyone who lives in Utah, whether you live in Logan or St. George.”
For example, a layer of dry low-density snow that receives a layer of wet heavy snow on top can be a recipe for dangerous avalanche conditions. So if his team has more confidence about when and where those conditions appear, they can close the right highways at the right times.
“People tend to think about that only when the road is closed,” Clark said. “What people tend to forget about is that we're actually able to keep the road open much more when we have really accurate weather forecasts.”
It could also help his team with the timing of snowplow crews and inform when road maintenance crews should be working or not.
The potential impacts of the forecast model highlight how funding for science can trickle down to improve people’s lives, Veals said. That’s been a concern in 2025 as the Trump administration has cut budgets for grants and programs nationwide.
“The past year has been a big year for Americans that seem to be questioning or unconvinced of where their federal tax dollars have gone in terms of science,” Veals said. “This is a great example of your tax dollars at work.”
The money that has made the snow forecast research possible came from the federal budget through a grant, he said, “and now all of us as Americans are going to benefit from these slightly improved snowfall forecasts.”
Editor’s note: KUER is a licensee of the University of Utah but operates as an editorially independent news organization.
