More Precise Local Weather Forecasts Enabled by AI

Machine learning and artificial intelligence (AI) are changing the way we predict weather and climate. NORCE is investing heavily in this field and has funded the project AIGLE – Artificial Intelligence Weather Prediction: Generalisation to Local Scales and Extremes through the research fund of NORCE Holding. The project is led by NORCE researcher Sigrid Passano Hellan, who is also affiliated with the Bjerknes Centre for Climate Research.

– Experience shows that machine learning performs well in weather prediction, with strong research communities now emerging in Norway and across Europe, says Hellan.

What is AIGLE?

The AIGLE project aims to position NORCE as a leading research centre for the use and evaluation of artificial intelligence in weather and climate modelling, with a particular focus on local conditions and extreme weather.

– We are building internal expertise within NORCE Climate, while also collaborating with other groups in NORCE that already have extensive experience with machine learning, including in Earth observation and Digital systems. It’s about preparing ourselves well, both academically and strategically, Hellan explains.

From Coarse to Detailed: Downscaling Weather Data

A key component of AIGLE is downscaling—moving from coarse, large-scale weather models to high‑resolution data that make sense at the local level. As part of the project, researchers are working, among other things, on downscaling weather data for southern Norway.

– We move from low resolution to high resolution so that we can provide more accurate information about the weather where people actually are, Hellan explains.

Small-scale, detailed weather data are particularly in demand from industry:

– For example, energy companies may want to know exactly how much wind will occur at the precise location of a wind farm. If we succeed in developing a downscaling tool that performs well and is cheaper than today’s solutions, the demand will be significant.

Testing Established Models — and Developing New Ones

Major technology companies have already invested heavily in AI-based weather prediction. Google, Microsoft, and IBM are all developing their own models, and there is also a great deal of activity in this field in Norway.

Recently, Hellan and research colleague Olav Ersland from the Norwegian Meteorological Institute organized a workshop on artificial intelligence and machine learning for weather forecasting. One of the highlights of the workshop was the presentation of Bris, a new Norwegian AI model for weather prediction in the Nordic region.

– In AIGLE, we have compared different models to see how they perform under Norwegian conditions and what their limitations are. We have also run a model ourselves to better understand how it works, says Hellan.

At the same time, NORCE has developed its own methods for high-resolution downscaling of weather and climate data over Norway.