Will machines soon make weather services superfluous?

Private offerings such as "Graphcast" from Google want to compete with authorities such as the German Weather Service. In view of the huge amounts of data, it seems logical that AI has advantages.

, and Ksenia Johnson

2024 January 4 . 4:53 AM

2 min read

Weather - Will machines soon make weather services superfluous?

Faster, more accurate, cheaper - this is how Google advertises its "Graphcast" product. Behind it is an artificial intelligence (AI). The AI model is able to "create medium-term weather forecasts with unprecedented accuracy", enthuses Remi Lam from the "Graphcast" research team.

"GraphCast" is not only faster, it can also warn of extreme weather events earlier, says Remi Lam. "It can predict the tracks of hurricanes in the future with great accuracy, identify atmospheric rivers associated with flood risk and predict the onset of extreme temperatures. This capability has the potential to save lives through better preparation."

DWD remains calm

In November, the Google researchers presented a comparison in the scientific journal "Science": according to the article, their AI predicted hundreds of weather variables over a ten-day period worldwide in less than a minute. For 90 percent of the metrics - such as temperature, wind speed or humidity - "Graphcast" performed better than the forecasts of the European Center for Medium-Range Weather Forecasts (ECMWF).

The German Weather Service (DWD) takes a critical view of such announcements - and remains calm. Yes, AI has "incredible potential", says meteorologist Andreas Walter, an expert in climate models at the DWD. AI may be faster, but it is by no means better. He sees the biggest deficits when it comes to predicting extremes that have not yet occurred.

This is due to the way the machine works. "Conventional numerical weather predictions use increased computing resources to improve forecast accuracy," explain the developers of Graphcast in Science. "However, they did not use historical weather data to improve the underlying model.

It is precisely this "reanalysis data", which is used to train the AI, that Andreas Walter believes is the problem: "The AI derives its learning algorithms from the past. Our models solve the basic physical equations."

Race for the best weather forecast

According to the DWD, once the initial state of the atmosphere has been recorded, in which all observation data is incorporated into the weather model, the equations are projected into the future in order to determine the future weather state. This numerical method is also used, for example, to create precipitation forecasts in the current flood situation. "Of course, this is a completely different effort. But it is also more reliable than an AI process that is only based on similarities," says Walter.

Speed is not really the point of forecasts, adds DWD spokesperson Uwe Kirsche. "It's not a race to produce the fastest weather forecast. It always has to be a race for the best weather forecast." Twice a day, the DWD calculates a global model for seven days in advance for 90 layers in the atmosphere. That takes about an hour. In addition, there are four runs a day for Europe and eight for Germany.

One limiting factor is computer performance. "Meteorologists need and receive more and more data. So we need ever larger computers that can also process this data," says Kirsche. The DWD will need a new mainframe computer in around two years' time. The current one cost around 120 million euros, and the next one is unlikely to be any cheaper.

AI is "certainly a tool that can be used to support meteorology", says Kirsche. The weather service is therefore currently testing AI in almost all areas. "Our goal is to improve the entire process chain - from data collection to delivery to customers - using AI," says Kirsche and emphasizes: "Not to replace: to improve."