ߣÏÈÉú

What will this winter’s snowfall totals in Colorado look like? It’s anybody’s guess, according to Joel Gratz, one of the most trusted names in snow forecasting.

A full crowd gathered in the lobby of the Westin Riverfront in Avon on Wednesday evening to hear from Gratz, the founder of the weather forecasting app OpenSnow. He explained, from a meteorological perspective, why this winter’s weather has forecasters guessing, and provided some tips on how to be more certain.

Gratz discussed last year’s weather, his predictions for this winter and how artificial intelligence will impact the future of weather forecasting.

---

---

Last winter

Last winter, an El Niño year, the entire West had below-average snowpack in December, and roughly average snowpack by the end of the season, with more snow to the south and less snow to the north. When assessing the “average,” meteorologists take a 30-year average, which generally runs from 1991 through 2020, Gratz said.

While an average snow year may be disappointing to powder chasers, “average is pretty good,” because it ensures adequate water resources, and not every year can see dramatic snowpack levels, Gratz said.

---

---

Another positive aspect of last winter’s snowfall was the consistency with which snow fell throughout the winter. 

This year and the meaning of La Niña

There has been very little snow so far this year. By this time last year, Eagle County had experienced its first snowfall. However, snow is expected to fall this weekend above 10,000 feet. The “best chance here” for snow is Sunday and Monday, Gratz said.

Snow will come again between Oct. 26 and 31, according to Gratz’s predictions.

“The beginning of the season is a crapshoot,” Gratz said. In other words, the amount of snow that falls from September through November is largely irrelevant in the grand scheme of winter in Colorado, having minimal effects on the overall snowpack and indicating little about the winter to come.

There is a 70-80% chance of a La Niña weather pattern this year.

La Niña is the name for when the waters of the central and east-central equatorial Pacific Ocean are colder than average. (By contrast, in an El Niño year, like last year, the same waters are warmer than average.)

This affects weather because the water temperature changes the path of the jet stream, causing it to move northward. This typically causes warmer, cooler winters in the northern regions of the United States and warmer, drier conditions in the South.

Looking at a map of the United States, Colorado lies right in the center of the La Niña weather pattern split, making it challenging for meteorologists to predict the impact on the state’s weather.

Cross-referencing the year’s expected La Niña variation with the historical data can sometimes help. However, of the six past seasons with similar La Niña patterns to this year, two saw above-average precipitation, two saw below-average precipitation and two saw average precipitation, Gratz said. One historical consistency is that December and January often saw above-average snowfall during La Niña years.

In other words, at this point, winter weather conditions in Colorado are near impossible to predict, Gratz said. (Gratz told the crowd that he shares the same information with his family and friends that he shared with them — no secrets.)

Knowing the year’s Atlantic Ocean temperatures can help predict precipitation levels, but even those numbers see a split between more and less precipitation that is draped across Colorado.

But weather forecasters and their tools are not completely irrelevant. The most accurate way to get a weather prediction for this winter is to begin looking at the forecast 10 to 15 days out, Gratz said.

The most accurate forecast, however, will always be the one closest to the day-of — and that forecast can be remarkably accurate. It’s just not possible to plan a ski vacation in October for February with any kind of certainty this year, with perhaps one exception — in January and February, Northern Japan has, statistically, the best chance of high snowfall and the most powder days, Gratz said.

Weather and AI

Artificial intelligence is assisting meteorologists in making great strides toward accuracy in their weather reporting by providing more data points much more rapidly, Gratz said.

AI weather models utilize pattern matching rather than running through an entire mathematical model, which takes seconds rather than several hours, Gratz said.

“They might have similar accuracy, but if you can run this (AI) in a second, and this one (mathematical model) takes six hours to run, you could run this one (the AI model) a thousand times, or a million times and get a bunch of different answers and try to figure it out,” he said.

AI models run 100 times were able to accurately predict the path of Hurricane Milton four to seven days in advance, and to predict the formation of Hurricane Helene two weeks before the storm made landfall, Gratz said.

At OpenSnow, forecasters run both high- and low-resolution models. Low-resolution models can be run faster and use fewer resources but are unable to take into account all the features of the terrain, like mountains, a key difference in Colorado. Using AI, low-resolution models can be trained to adjust to the information on the high-resolution models. This can improve the accuracy of the low-resolution models by 10 to 30%.

“The fact that we can do this as a 10-person company with two people working on this for a couple of months is so cool,” Gratz said.