Judah Cohen, a research scientist at MIT, is utilizing artificial intelligence to bridge the gap in subseasonal forecasting, specifically targeting the two-to-six-week range that has historically challenged meteorologists. While traditional winter forecasts often depend heavily on the El Niño–Southern Oscillation, Cohen’s innovative approach shifts the focus toward high-latitude diagnostics. By integrating data points such as Siberian snow cover, sea-ice extent, and polar vortex stability, the model identifies precursors to extreme weather that standard models often overlook.

The efficacy of this methodology was demonstrated during the 2025 AI WeatherQuest competition, where Cohen’s team secured first place by outperforming statistical baselines and conventional AI weather models. The system employs sophisticated machine-learning pattern recognition to interpret complex atmospheric interactions within the Arctic circle. This breakthrough allows for the identification of subtle signals that dictate broader hemispheric temperature shifts. Unlike models relying on tropical influences, this AI-driven tool provides a more nuanced understanding of northern climate dynamics.

Practical applications were evident this season when the model successfully flagged a significant mid-December cold surge for the U.S. East Coast weeks before traditional signals emerged. Even during periods of weak ENSO activity, the system’s ability to decode Arctic indicators provides critical lead time for public safety and energy infrastructure planning. By combining specialized Arctic data with machine learning, researchers provide utilities and transportation agencies with the foresight needed for effective preparation. This advancement represents a major step forward in mitigating the impacts of extreme winter events.

Judah Cohen, a research scientist at MIT, is utilizing artificial intelligence to bridge the gap in subseasonal forecasting, specifically targeting the two-to-six-week range that has historically challenged meteorologists. While traditional winter forecasts often depend heavily on the El Niño–Southern Oscillation, Cohen’s innovative approach shifts the focus toward high-latitude diagnostics. By integrating data points such as Siberian snow cover, sea-ice extent, and polar vortex stability, the model identifies precursors to extreme weather that standard models often overlook.

The efficacy of this methodology was demonstrated during the 2025 AI WeatherQuest competition, where Cohen’s team secured first place by outperforming statistical baselines and conventional AI weather models. The system employs sophisticated machine-learning pattern recognition to interpret complex atmospheric interactions within the Arctic circle. This breakthrough allows for the identification of subtle signals that dictate broader hemispheric temperature shifts. Unlike models relying on tropical influences, this AI-driven tool provides a more nuanced understanding of northern climate dynamics.

Practical applications were evident this season when the model successfully flagged a significant mid-December cold surge for the U.S. East Coast weeks before traditional signals emerged. Even during periods of weak ENSO activity, the system’s ability to decode Arctic indicators provides critical lead time for public safety and energy infrastructure planning. By combining specialized Arctic data with machine learning, researchers provide utilities and transportation agencies with the foresight needed for effective preparation. This advancement represents a major step forward in mitigating the impacts of extreme winter events.