Czech meteorologists to use AI for enhanced weather forecasting

The future of weather forecasting will benefit from the collaboration between meteorologists and artificial intelligence (AI), Radek Tomšů, director of forecasting services at the Czech Hydrometeorological Institute (ČHMÚ), tells Czech News Agency.

Tomšů explained that meteorologists are currently utilizing machine learning technologies, specifically AI models, to enhance predictions for specific weather phenomena such as fog. He emphasizes that improved forecasting will be a result of experts working with these new systems, not being replaced by them.

Recently, Google DeepMind developed the GraphCast model, which promises accurate forecasts up to ten days in advance. GraphCast employs graph neural networks (GNNs) to deliver global weather forecasts, with the potential to surpass existing numerical prediction methods.

Tomšů cautioned that the model requires further testing before it can be implemented at the ČHMÚ, and that the development process must undergo rigorous long-term benchmarking tests to ensure reliability.

“Machine learning and deep neural networks identify patterns in the data that traditional models often fail to capture,” Tomšů says. “Initial results are promising, but we still need to verify how reliable this model is and where its weaknesses may lie.”

Currently, GraphCast is being tested at the European Centre for Medium-Range Weather Forecasts (ECMWF). It offers advanced forecasting capabilities, but Tomšů points out that it focuses on a limited range of meteorological elements.

“We need broader coverage and more comparisons through testing to know if this technology development is relevant to us or if we are going down a blind path,” he cautioned.

One of GraphCast’s key advantages is its speed; it can generate a ten-day forecast in just one minute, while traditional numerical models require several hours for similar calculations.

In addition to surface variables such as temperature and wind speed, the new model processes atmospheric variables at 37 altitude levels, aiming to cover as many meteorological elements as possible. However, research is still in the early stages.

In a recent meeting, experts from the ČHMÚ agreed to plan further expansion and integration of AI technologies into their workflows. “In the future, we will be even more involved in artificial intelligence in our scientific work,” Tomšů notes.

As AI technology continues to evolve, he envisions a future where meteorologists and artificial intelligence work together to provide more accurate and timely weather forecasts for society. This collaboration could enhance preparedness for extreme weather events such as the recent floods that devastated Central Europe.