AI model transforms weather forecasting with near-instant results

In 2020, researchers at the University of Washington in the United States sought to improve weather forecasting methods that were traditionally reliant on numerical weather prediction (NWP), which has been the industry standard for decades. This older approach involves complex mathematical computations to replicate real-world physics, a process that is both slow and resource-intensive, often requiring significant computing power and expert interpretation. Despite its long-standing use, NWP was often inaccurate in its predictions, prompting a need for innovation in the field of meteorology. Enter the AI approach. The team developed a statistical model that leverages machine learning techniques similar to those used in systems like ChatGPT. This new model harnesses decades' worth of historical weather data to predict future patterns based on real-time weather conditions. Lead researcher Jonathan Weyn noted that the AI's capabilities allow it to recognize typical patterns in weather and extrapolate them rapidly, with remarkable efficiency and accuracy. By 2023, advancements made in AI enhanced the reliability of these models, with some outperforming traditional methods over shorter forecasting periods. The introduction of AI technology in weather forecasting has led to breakthrough innovations like the Aardvark model from researchers at Cambridge, unveiled in March 2023. This model can deliver more accurate forecasts using only a fraction of the data input required by existing systems, showcasing a significant democratization of weather prediction. By making these advanced forecasting technologies accessible on standard desktop computers, researchers aim to empower developing nations and regions with limited data access, contributing to a more equitable approach to climate governance. Most notably, DeepMind’s GraphCast model successfully tracked Hurricane Lee nine days prior to its landfall—well ahead of traditional forecasts, which lagged by an additional three days. The implications of this are vast, with AI models also contributing valuable insights into other phenomena, such as wildfires, through pattern recognition of environmental factors. Notably, major organizations like the National Oceanic and Atmospheric Administration (NOAA) and the European Centre for Medium-Range Weather Forecasts (ECMWF) have integrated these AI systems into their forecasting processes, with collaborations established with institutions like the Alan Turing Institute to further refine these capabilities. As the AI revolution unfolds, its timing becomes crucial in addressing the mounting environmental challenges that societies face today.