{"id":22451,"date":"2025-03-26T19:00:00","date_gmt":"2025-03-26T17:00:00","guid":{"rendered":"https:\/\/forklog.com\/en\/scientists-develop-ai-model-for-accurate-weather-forecasting\/"},"modified":"2025-03-26T19:00:00","modified_gmt":"2025-03-26T17:00:00","slug":"scientists-develop-ai-model-for-accurate-weather-forecasting","status":"publish","type":"post","link":"https:\/\/forklog.com\/en\/scientists-develop-ai-model-for-accurate-weather-forecasting\/","title":{"rendered":"Scientists Develop AI Model for Accurate Weather Forecasting"},"content":{"rendered":"

Researchers from the UK and Canada have developed the AI model Aardvark Weather for weather forecasting. The results are presented in a report in Nature<\/a>.<\/p>\n

The traditional approach to prediction involves modeling the physics of the atmosphere using complex equations. Scientists from the University of Cambridge, the Vector Institute at the University of Toronto, and the Alan Turing Institute have created a tool for “deep learning”\u2014training computers to recognize patterns in large volumes of data.<\/p>\n

The model is designed to generate global forecasts for wind, humidity, geopotential<\/span>, and temperature. <\/p>\n

\n

\u201cCurrently, the forecasting system has several components that require significant computational resources. We have been able to replace many of these labor-intensive elements with much lighter models trained to perform the same tasks,\u201d commented<\/a> one of the researchers, James Requena.<\/p>\n<\/blockquote>\n

As a result, Aardvark Weather makes forecasts more frequently and with better speed and accuracy. The team replaced each step in the prediction pipeline, including the transformation of raw data.<\/p>\n

\n

\u201cWe found that when these machine learning components are linked in a chain, the overall performance significantly improves. Fine-tuning the entire pipeline for the final task we set ourselves allows us to optimize each element not only for its individual role but also for how it contributes to achieving the result we care about most,\u201d noted Requena.<\/p>\n<\/blockquote>\n

Aardvark Weather uses raw atmospheric data\u2014measurements of pressure, temperature, and humidity\u2014to produce high-quality global and local forecasts.<\/p>\n

The system is built on three neural components: an encoder, a processor, and a decoder.<\/p>\n