{
  "id": 14475925,
  "title": "AI-enhanced weather model doubles predictive power two to six weeks out",
  "url": "https://urgent.news/2026/10/14/ai-enhanced-weather-model-doubles-predictive-power-two-to-six-weeks",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-14T15:00:47.000Z",
  "source": {
    "name": "Phys.org",
    "slug": "phys-org",
    "url": "https://phys.org/news/2026-10-ai-weather-power-weeks.html"
  },
  "original_language": "en",
  "account": "In 2021, Texas endured the most severe energy infrastructure failure in state history following three intense winter storms in February, resulting in at least 246 deaths due to shortages of food, water, heat, and electricity. Energy officials could have mitigated the damage by having more time to prepare, such as bringing in more trucks of natural gas to disrupted regions. Weather forecasts beyond two weeks are notoriously unreliable, leaving a limited window for energy operators, farmers, and emergency response officials worldwide.\n\nA team of researchers, including those from MIT, developed a machine-learning model that learns to correct errors in existing forecasting models to enhance their output over the two-to-six-week period. When tested against a benchmark forecasting model, this new technique doubled the forecasting accuracy over that time frame and secured the first place in a prominent international forecasting competition. In a study published in Nature, the researchers describe their method and demonstrate how it can generate forecasts that surpass climatologic averages for the first time between two and six weeks.\n\nThis improvement not only benefits time-averaged forecasts for farmers and energy industry professionals, who aim to align supply with demand, but also enhances predictions of extreme weather events, including flooding, cold spells, and other phenomena. Co-author Judah Cohen, an MIT research scientist, explains the significance of accurately predicting both routine and extreme weather events, as the latter can have a substantial societal impact.\n\nWhile traditional weather forecasting is limited by a two-week range, known as the \"predictability desert,\" due to the chaotic nature of the atmosphere, the researchers believe their approach could help overcome this limitation. In the 1970s, MIT professor Edward Lorenz introduced the concept of chaos theory to explain how minor variations in initial weather conditions can lead to vastly different outcomes beyond this two-week period. Cohen contends that while the atmosphere may theoretically generate an infinite number of patterns, it often tends to fall into familiar patterns repeatedly. By recognizing and predicting these recurring patterns, it may be possible to anticipate subsequent events.\n\nCurrently, the most advanced weather forecasts are produced through dynamical models, which numerically simulate the atmosphere and ocean's evolution over time, or artificial intelligence models that learn to make predictions from historical weather data. AI models have demonstrated improvements in short-range predictions but have been less effective for longer time ranges. Cohen's research focuses on subseasonal forecasting, which spans two to six weeks and falls within the predictability desert. This domain is characterized by the absence of both short-term initial conditions and long-term slow-varying boundary forces, making it difficult to forecast accurately.",
  "summary": "In 2021, Texas suffered the worst energy infrastructure failure in state history, following three severe winter storms in February. At least 246 people were killed as a result of the disruptions, which led to food, water, heat and electricity shortages.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}