Last weekend was a mix between lovely beach weather and heavy rain in the Netherlands. So an accurate weather forecast is more than welcome. AI models trained with millions of hours of historical weather data are now delivering predictions as precise as those from supercomputers, according to an article in the scientific journal Science.
Why this is important:
Weather forecasts are getting better and better. Where a six-day forecast is now as reliable as a three-day forecast 30 years ago, AI is providing even greater accuracy and speed. Thanks to deep learning and pattern recognition, AI models are able to predict changes in the atmosphere accurately.
Sudden changes
AI weather models, such as Google DeepMind’s GraphCast, can forecast up to 10 days ahead. These can generate accurate forecasts with less computing power and in a fraction of the time, using data from the past 40 years. This is especially important for sudden changes in weather, such as the course of a hurricane.
Big tech companies
Major technology companies such as Google, Microsoft, and Huawei are actively contributing to the development of these advanced AI weather models. Google DeepMind and Huawei are competing for the most proficient AI weather model, with GraphCast outperforming European weather forecasting organization ECMWF on 90% of verification targets. This competition accelerates progress and provides ever-better and faster weather forecasts.
In the Netherlands, AI companies are also active in the field of weather forecasting. Beyond Weather uses advanced AI to provide accurate weather forecasts for the energy and agriculture sectors across Europe.
Human factor
Although AI is capable of generating weather reports quickly and accurately, the role of the meteorologist remains indispensable. Meteorologists bring intuition and experience that are essential for assessing and communicating risk. AI can automatically produce forecasts, but interpreting and explaining this data to the public remains human work.
Black box
The adoption of AI in weather forecasting is promising but also has challenges. One of the biggest obstacles is the “black box” nature of AI models; it is often unclear how these systems arrive at their conclusions. This can lead to hesitation among users and researchers. Still, as long as predictions are accurate, the expectation is that the acceptance of AI in this sector will continue to grow.