Google's DeepMind Develops AI for Enhanced Weather Forecasts

Google subsidiary DeepMind has made significant strides in weather forecasting with the development of an artificial intelligence (AI) model.

Credit: Jam Sta Rosa/AFP/Getty Images

The model, called GraphCast, utilises decades of historical weather data to generate highly accurate 10-day forecasts in less than a minute.

In a study published in the journal Science, GraphCast outperformed the European Centre for Medium-Range Weather Forecasts' gold-standard forecasting model on approximately 90% of tested metrics.

The AI model not only excelled in general weather predictions but also demonstrated proficiency in forecasting extreme weather events. It has the capability to predict severe temperatures and track the paths of cyclones, enabling more effective preparation and response to potentially devastating weather conditions. By leveraging vast amounts of weather data and real-time conditions, GraphCast identifies patterns and provides forecasts that extend further into the future than previously possible.

Rémi Lam, a research scientist at DeepMind, emphasised that GraphCast should be seen as a complementary tool to conventional forecasting models rather than a replacement. The AI-based system has the potential to enhance existing methods and support decision-making processes crucial to various industries and societies.

---

• Google's DeepMind has developed an AI model, GraphCast, for more accurate weather forecasts.

• GraphCast utilises historical weather data and current conditions to generate 10-day forecasts in under a minute.

• The AI model outperformed the European Centre for Medium-Range Weather Forecasts' gold-standard model on approximately 90% of tested metrics.

Source: SCMP