Every autumn, as the earth’s Northern Hemisphere tilts away from the sun, ushering in shorter days, Judah Cohen, a seasoned research scientist at MIT’s Department of Civil and Environmental Engineering, is set to solve a formidable atmospheric puzzle. The goal behind his extensive research is primarily to decipher the way Arctic conditions shape winter weather patterns across Europe, Asia, and North America.
Cohen’s career, which took off in earnest during his postdoctoral research with Professor Dara Entekhabi, commenced with a focus on Siberian snow cover. Today, his study branches out into an encompassing investigation of high-latitude climate signals and their role in molding the winter season. With the onset of artificial intelligence, Cohen’s predictive weather analyses are reaching unprecedented heights. But the story doesn’t end there.
Winter forecasts of old have typically depended heavily on the El Niño–Southern Oscillation (ENSO), a climate trend driven by tropical Pacific Ocean temperatures. However, the twist in the tale is the weaker ENSO this year, which results in minimal climatic influence, nudging Cohen to look to the Arctic for answers.
“When ENSO is weak, Arctic climate indicators play a pivotal role,” Cohen asserts. He vigilantly observes Arctic parameter, including October snow cover in Siberia, early-season temperature anomalies, and sea-ice extent, along with the behavior of the polar vortex. Making sense of these factors, he explains, can shed surprising light on the coming winter.
Just last October, while a large part of the Northern Hemisphere experienced unusually warm temperatures, Siberia proved to be an exception. With temperatures lower than normal and early snowfall, it shaped conditions traditionally leading to the formation of cold air masses. These chilling air masses could shift into Europe and North America, introducing more regular cold spells.
Factors such as warm ocean temperatures in the Barents–Kara Sea and a specific phase of the quasi-biennial oscillation indicate that the polar vortex might be more unstable than usual. If this aligns with certain surface conditions in December, it could cause below-normal temperatures across vast swaths of Eurasia and North America earlier in the winter season.
Where AI has already marked significant strides are in short-range weather predictions, stretching from one to ten days. Yet, applying AI to lengthier timeframes has remained challenging. Specifically, subseasonal forecasting covering two to six weeks has been meteorology’s long-standing hurdle. But, it seems, times are changing.
This year, Cohen and his squad were crowned winners in the fall season of the 2025 AI WeatherQuest subseasonal forecasting competition, organized by the European Centre for Medium-Range Weather Forecasts (ECMWF). The contest evaluated how effectively AI models could predict temperature patterns weeks ahead. The winning model, which married machine learning with Cohen’s extensive Arctic climate experience, handily outperformed incumbent AI and statistical baselines.
“This level of performance, if consistent across multiple seasons, could mark a massive stride towards enhancing subseasonal prediction,” notes Cohen. Furthermore, the model’s detection of a potential cold surge along the U.S. East Coast in mid-December, weeks prior to conventional models capturing it, drew considerable media attention. If validated, this could signal a significant enhancement in the early warning capabilities for extreme weather events.
As per Cohen’s new model, the winter of 2025-26 might bring below-average temperatures to areas of Eurasia and central North America, particularly during midwinter. While these are early predictions subject to shifts, the fundamental signals for a colder winter appear to be in place.
As the Arctic keeps warming up at an alarming rate, its impact on winter weather grows significantly pronounced. Grasping these connections is essential for planning in large sectors, including energy, transportation, and public safety.
Cohen firmly believes that the Arctic brims with untapped potential to improve subseasonal forecasts, and AI could be the key to unlocking this. His work is inching steadily towards more than just the scientific frontier—it’s also paving the way for broader cultural recognition. A testament to the rising popularity of his research was when Cohen appeared as a clue in The Washington Post crossword puzzle in November.
“The Arctic, for me, has always been a key area to observe. Now, AI is providing new avenues to interpret its signals,” says Cohen. To keep up with Cohen’s ongoing winter outlook, you can check out his blog.
Oryginalne źródło: https://news.mit.edu/2026/decoding-arctic-to-predict-winter-weather-0108
This website uses cookies.