{
  "id": 10095496,
  "title": "Why AI has trouble predicting the fury of hurricane intensity",
  "url": "https://urgent.news/2026/09/27/why-ai-has-trouble-predicting-the-fury-of-hurricane-intensity",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-27T00:30:01.000Z",
  "source": {
    "name": "Phys.org",
    "slug": "phys-org",
    "url": "https://phys.org/news/2026-09-ai-fury-hurricane-intensity.html"
  },
  "original_language": "en",
  "account": "Artificial intelligence has taken meteorology by storm, producing forecasts that match the best physics-based models. This progress is owed to three key elements: abundant weather data, advancements in AI models, and unprecedented computing power. However, while the focus remains on enhancing models or hardware, data remains the cornerstone. AI has leveraged decades of climate and weather records globally, enabling it to recognize patterns that would have been unattainable a decade ago. Yet, the transition from global to regional forecasting poses significant challenges for AI. This particularly holds true for predicting hurricane intensity, as rapid intensification can turn harmless storms into deadly behemoths within hours. Hurricane Polo, for instance, intensified from a tropical storm to a Category 5 hurricane in just 24 hours off Mexico's Pacific coast. Such rapid intensification poses a major challenge for forecasters, leaving communities with inadequate time to evacuate and prepare. Unlike global forecasts, hurricane intensity predictions are a regional issue, with extreme events often developing rapidly or moving swiftly over short periods. Regional forecasts concentrate on phenomena like heavy rainfall, squall lines, severe thunderstorms, or hurricanes. Capturing these behaviors in AI models necessitates data at a finer resolution than present global datasets can offer. During model training, scientists typically use two datasets. The first is observations, comprising measurements of rainfall, near-surface temperature, wind speed, and other weather variables from weather stations, radars, buoys, and satellites. While these observations are detailed, they are often limited to coastal areas and unevenly distributed. Many critical stages of hurricane development occur over the open ocean, where direct observations are scarce. Satellites can partially fill these gaps in the open ocean, but their coverage is limited, making it challenging to estimate rainfall, wind speeds, or cloud-top temperatures comprehensively. The second dataset comprises weather model simulations, which provide the most complete three-dimensional depiction of the atmosphere at high resolution by combining atmospheric conditions and physical knowledge. However, these simulations are not flawless either, as all computer models incorporate approximations and uncertainties stemming from incomplete knowledge of Earth's atmosphere. Thus, there are always fine-scale processes that simulations cannot capture. Consequently, we lack a comprehensive, full three-dimensional dataset to train AI models for hurricane intensity prediction at present. While more detailed data would be ideal, it is not the only hurdle for AI in predicting hurricane intensity. Even if scientists could gather precise measurements of every aspect of thousands of storms worldwide every second, AI might still struggle to predict intensity accurately. This is due to chaos - the inherent unpredictability within hurricanes. My research with colleagues suggests that hurricanes may possess a form of chaos that hinders AI's ability to forecast intensity over long periods. Once a tropical storm embeds itself in a favorable environment, it can intensify to its maximum potential strength, known as the potential intensity. This limit is primarily determined by the surrounding environment, such as warm ocean water fueling hurricane intensity and wind shear slowing development. Any minor disturbances will cause fluctuations in hurricane intensity, which tend to be greater when the ocean surface is warmer. Recent studies propose that these fluctuations are not random but occur within a chaotic attractor—a set of possible storm states within which the hurricane can evolve unpredictably. Although the existence of such a chaotic intensity attractor has not been conclusively established, it poses a fundamental challenge for training AI models to predict hurricane intensity accurately. While scientists strive for the most accurate predictions, capturing the hurricane's intrinsic chaos is equally important. However, if an AI model captures this chaos, its error cannot be reduced indefinitely. An AI model trained to minimize forecast error may, instead, learn the most likely evolution of a hurricane while smoothing out uncertainties and unpredictability.",
  "summary": "Artificial intelligence has revolutionized weather forecasting in just a few years, with global AI weather models now able to produce forecasts that rival some of the world's best physics-based prediction systems.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 3,
    "also_reported_by": [
      {
        "outlet": "The Conversation",
        "title": "Why AI has trouble predicting the fury of hurricane intensity",
        "url": "https://urgent.news/2026/09/23/why-ai-has-trouble-predicting-the-fury-of-hurricane-intensity",
        "published": "2026-09-23T12:44:25.000Z"
      },
      {
        "outlet": "Jamaica Gleaner",
        "title": "Why AI has trouble predicting hurricane intensity",
        "url": "https://urgent.news/2026/09/26/why-ai-has-trouble-predicting-hurricane-intensity",
        "published": "2026-09-26T05:07:19.000Z"
      }
    ]
  },
  "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."
}