Every autumn, as the Northern Hemisphere heads toward winter, Judah Cohen begins to piece together a complex atmospheric puzzle. Cohen, a research scientist in MIT’s Department of Civil and Environmental Engineering (CEE), has spent decades studying how Arctic conditions determine the course of winter weather across Europe, Asia, and North America. Her research is linked to her postdoctoral work with Bacardi and Professor Dara Entekhabi of the Stockholm Water Foundation, which looked at snow cover in the Siberian region and its relationship with winter forecasting.
Cohen’s outlook for the winter of 2025-26 highlights a season in which indicators emerging from the Arctic use a new generation of artificial intelligence tools that help develop complete atmospheric pictures.
Looking beyond the usual climate drivers
Winter forecasts depend heavily on El Niño–Southern Oscillation (ENSO) diagnostics, which are conditions in the tropical Pacific Ocean and atmosphere that influence weather around the world. However, Cohen says ENSO is relatively weak this year.
“When ENSO is weak, climate indicators from the Arctic become especially important,” says Cohen.
Cohen tracks high-latitude dynamics in his subseasonal forecast, such as October snow cover in Siberia, early-season temperature changes, Arctic sea-ice extent, and the stability of the polar vortex. “These indicators can tell a surprisingly detailed story about the upcoming winter,” he says.
One of Cohen’s most consistent data predictors is October weather in Siberia. This year, while October in the Northern Hemisphere was unusually warm, Siberia was colder than usual due to early snowfall. “The colder temperatures combined with early snow cover strengthen the formation of cold air masses that can later spread over Europe and North America,” says Cohen, “weather patterns that have historically been associated with greater cold later in the winter.”
Warm sea temperatures in the Barents–Kara Sea and the “eastern” phase of the semi-biennial oscillation also suggest a potentially weak polar vortex at the beginning of winter. When this disturbance combines with surface conditions in December, it causes below normal temperatures over parts of Eurasia and North America early in the season.
AI subseasonal forecast
While AI weather models have made impressive progress in short-range (one to 10 day) forecasts, these advances have not yet been applied to longer ranges. Sub-seasonal forecasting ranging from two to six weeks remains one of the most difficult challenges in the region.
It is with this difference that this year could be a turning point for subseasonal weather forecasting. A team of researchers working with Cohen won first place for autumn weather in the 2025 AI WeatherQuest subseasonal forecast competition, organized by the European Center for Medium-Range Weather Forecasts (ECMWF). The challenge is to evaluate how well AI models capture temperature patterns over several weeks, where predictability has historically been limited.
The winning model combined machine-learning pattern recognition with the same arcane diagnostics that Cohen has refined over decades. The system demonstrated significant gains in multi-week forecasting, surpassing leading AI and statistical baselines.
“If this level of performance is sustained over multiple seasons, it could be a real step forward for sub-seasonal prediction,” says Cohen.
The model also detected a potential cold surge much earlier than normal along the US East Coast in mid-December, several weeks before such signs usually occur. The forecast was widely publicized in the media in real time. If validated, Cohen explains, it will show how combining Arctic indicators with AI can effectively increase lead times for weather predictions.
“Marking any potential extreme event three to four weeks in advance would be an important moment,” he added. “This will give utilities, transportation systems and public agencies more time to prepare.”
what can happen this winter
Cohen’s model shows the potential for colder than normal temperatures over parts of Eurasia and central North America in late winter, with the strongest anomalies likely mid-season.
“We’re still in the early stages and the pattern could change,” Cohen says. “But the ingredients are in place for a colder winter pattern.”
As Arctic warming accelerates, its effects on winter behavior are becoming more apparent, making it important to understand these connections for energy planning, transportation, and public safety. Cohen’s work shows that the Arctic holds untapped subseasonal forecasting power, and AI can help unlock it for time ranges that have long been challenging for traditional models.
In November, Cohen also came forward as a clue Washington Post The crossword is a small indication of how widely their research has entered the public conversation about winter weather.
“For me, the Arctic has always been a place worth seeing,” he says. “Now AI is giving us new ways to interpret its signals.”
Cohen will continue to update his outlook throughout the season on his blog.
