Better weather forecast thanks to AI

Weather forecasts are often unreliable. But to reduce the impact of extreme weather, such as drought or heat, it is precisely important to know in time when extreme weather is coming. Van Straaten uses a combination of physical and statistical weather models for this purpose. With these, he shows that the distribution of cold and warm air can be better predicted when looking at longer periods and larger areas.  

 

Using machine learning techniques, Van Straaten and his fellow researchers discovered that processes in the Pacific Ocean are important in steering that distribution between cold and warm air. This link in the climate system was found to be missing in the model used for forecasting by many weather apps. They developed, in collaboration with KNMI, a method that corrects that model statistically and makes it more useful. 

 

This method also provides more insights into the errors of weather forecasts, including by comparing forecasts and actual measured weather. According to Van Straaten, the method is a clear way in which classical physics-based weather models can be improved with AI. 'All buyers of weather forecasts, from water boards, to farmers, to ordinary citizens, can benefit. They will have early access to reliable information'.

This article can be read on the Vrije Universiteit Amsterdam website.