A Breakthrough in Weather Forecasting
GraphCast, developed by Google DeepMind, has set a new standard in weather forecasting, winning the prestigious 2024 MacRobert Award. Trained on four decades of historical data, GraphCast can predict a wide range of variables such as temperature, wind speed, humidity, and pressure at multiple atmospheric levels, offering unprecedented accuracy, in under a minute.
Higher Performance and Applications
GraphCast outperforms traditional forecasting models, providing 10-day forecasts in under a minute using a single Google TPU v4 machine—significantly faster than the hours-long process required by supercomputers. It has already demonstrated its ability to predict extreme weather events, including accurately forecasting Hurricane Lee’s landfall in Nova Scotia three days earlier than traditional methods. With its ability to track cyclones, atmospheric rivers, and extreme temperature events, GraphCast is poised to transform how we approach weather forecasting.
Source: Google DeepMind
Future of Weather Forecasting
The model is already being adopted by the European Centre for Medium-Range Weather Forecasts (ECMWF) and has been open-sourced by Google DeepMind to allow access globally. With energy efficiency around 1,000 times better than traditional methods, GraphCast offers a more sustainable and reliable solution for predicting extreme weather events, enabling faster and more accurate decision-making globally.
