Step aside, Met Office! Google’s AI can accurately predict the weather forecast 15 DAYS in advance

Getting caught in the rain may soon be a thing of the past thanks to a powerful new AI weather forecaster.

Google DeepMind has unveiled an AI-powered weather model called GenCast that it claims is faster and more accurate than traditional forecasts.

Compared to the best-performing supercomputer, Google’s GenCast model was more accurate on 99.8 percent of predictions up to 15 days in advance.

According to Google, this will not only help commuters decide whether to bring an umbrella, but also spot natural disasters like typhoons before it’s too late.

Normally, weather agencies such as the Met Office predict the weather by using huge supercomputers to break through the complex mathematics that simulate the climate.

GenCast, on the other hand, uses AI to discover patterns in historical weather data and create fifty possible outcomes that form the basis of an ‘ensemble forecast’.

When the majority of these possibilities show the same weather events occurring, scientists can predict the weather with a high degree of confidence.

Google DeepMind engineer Ilan Price says: ‘Such forecasts are more useful than relying on a single forecast because they give decision makers a fuller picture of possible weather events in the coming days and weeks and how likely each scenario is.’

Because weather patterns are so complex, the best weather forecasts are “probabilistic,” meaning each outcome is assigned a probability of occurring.

While this provides a more complete picture of the weather, it also requires an enormous amount of time and computer power.

In a new paper published in NatureGoogle DeepMind shows that its new AI model is more accurate than the best-performing ENS model from the European Center for Medium-Range Weather Forecasts (ECMRWF).

Both GenCast and ENS are probabilistic, but the way they create their set of predictions is completely different.

Rather than trying to simulate the complex physics of the atmosphere, GenCast uses a type of AI called a diffusion model, which is commonly found in video, image and music generators.

When the AI ​​is provided with the most recent weather condition, it generates fifty predictions for the future weather condition, just as some AIs can create images via a text prompt.

The difference is that GenCast is specifically adapted to work on the spherical surface of the Earth and is trained on 40 years of weather data.

Google claims that this method is not only faster, but also provides a better forecast for both daily weather and extreme events than ENS.

The AI ​​was trained on data back to 2018 and then evaluated against real weather data from 2019 and ENS forecasts for that year.

GenCast was more accurate than ENS on 97.2 percent of predictions and on 99.8 percent when making predictions more than 36 hours in advance.

Most notably, when both systems were tasked with predicting the arrival of Typhoon Hagibis, GenCast was able to provide a warning 12 hours earlier.

When Typhoon Hagibis hit Japan in 2019, it was the worst storm in six decades and led to widespread destruction.

The engineers behind GenCast hope that by alerting authorities earlier, AI-powered weather forecasts can help save lives.

Mr Price said: ‘As climate change leads to more extreme weather events, accurate and reliable forecasts are more important than ever. However, the weather cannot be predicted perfectly and predictions are uncertain, especially after a few days.

“Getting better and more advanced warnings about where they will attack land is invaluable.”

Poor forecasts based on traditional methods have led to deadly consequences in the past when reports downplayed the dangers of approaching storms.

For example, in 1987, BBC weatherman Michael Fish assured viewers that no hurricane was heading for Britain.

The next day, devastating hurricane force winds hit Britain, killing 18 people and causing £1 billion ($1.3 billion) in damage.

However, the power of computational weather forecasting has come a long way since the 1980s.

When Michael Fish made his fateful prediction, the Met Office’s supercomputer had the processing power equivalent to that of an average smartphone today.

Currently, the Met Office has upgraded to the Cray XC40 supercomputer system, which can perform more than 14,000 trillion arithmetic operations per second.

Even the ENS model against which GenCast was measured has improved significantly in recent years.

In their paper, Google used DeepMind ENS 2019 predictions, but ECMRWF has since made some significant improvements.

In particular, ENS is now capable of producing significantly higher resolution predictions than GenCast can produce.

GenCast divides the world into a grid and looks at squares with a latitude and longitude of 0.25 degrees.

In contrast, the ENS weather forecast now works with a resolution of just 0.1 degrees, which means finer forecasts.

Google DeepMind admits that traditional models are likely to be irreplaceable for the foreseeable future – not least because they provide the data to train AI.

However, AI predictions have a major advantage in terms of speed and computing power.

Traditional predictions like the ENS take hours on a supercomputer with tens of thousands of processors.

In contrast, GenCast takes just eight minutes to produce a 15-day forecast using a single processing unit.

In the future, this means that AI models could become much more common for applications such as predicting extreme weather or planning around renewable energy sources such as solar and wind energy.