Could AI cost me my job as a weather forecaster?

Weather forecasts play a huge role in our daily lives. They help us to decide how we get to work, if we need an umbrella and when to hang out the washing.

For industries like farming, shipping, aviation and renewable energy, accurate forecasts are essential.

I've been presenting the weather on the BBC for nearly 30 years. Over that time, I've witnessed how our changing climate is bringing more extreme and intense weather events.

It is so important to keep improving the way we predict the weather, and scientists are exploring how AI could make forecasting more accurate, efficient, and faster than ever before.

In some countries, broadcasters have even begun to experiment with AI-generated weather presenters.

But does all this mean that forecasters like me could soon be out of a job?

Traditionally, weather forecasting relies on complex numerical weather prediction models which require vast amounts of data and supercomputers - like the ones used by the Met Office.

But the Met Office is now working with experts from the Alan Turing Institute, the UK's national centre for data science and AI, to build a new global forecasting system powered by AI.

One of their models, called FastNet, uses machine learning to improve prediction capabilities.

Prof Kirstine Dale, Chief AI Officer at the Met Office, said it had the potential to revolutionise forecasting.

"AI is phenomenally fast - not just a bit faster, but tens of thousands of times faster," she told me.

"That means it can produce up to date forecasts with a fraction of the computational cost and carbon dioxide."

She said AI could also produce "hyper-localised" forecasts, "potentially offering more up to date forecasts tailored to your postcode".

Forecasts driven by AI could also help to mitigate against the impacts of storms, floods and heatwaves by providing earlier and more accurate warnings of severe weather.

But there are challenges - especially in predicting rare or extreme weather events.

"The past is no longer a reliable indicator of the future," said Prof Dale.

"So we need traditional numerical weather prediction (NWP) models to explore how the climate may change and generate recalibrated datasets.

"These recalibrated datasets of future climates can be used to train AI-based models.

"AI-based models simply aren't aware of the physics - and changing physics - of the atmosphere, so NWP will likely continue to play a vital role in forecasting extreme events, as well as adding a layer of validation to AI forecasting outputs."

Dr Scott Hosking, Mission Director for Environmental Forecasting at the Turing Institute, said that once trained, AI models were cheaper and quicker to run than traditional forecasting systems.

"AI has surprised us in a number of ways," said Dr Hosking.

"One of them is how well it predicts the tracks of cyclones and hurricanes. AI is always learning what it has seen in the past."

But he said AI still had some way to go in certain areas, including in its ability to forecast high-intensity rainfall - the kind that often leads to flash flooding.

AI could also play a key role in space weather forecasting, helping to predict solar storms more accurately and efficiently.

These storms, caused by solar activity, are best known for producing the aurora borealis - the Northern Lights - which have been seen in Wales several times recently.

But space weather can be hazardous too, affecting Earth's magnetic field and potentially disrupting communication systems and infrastructure.

Dr Huw Morgan, Head of Solar System Physics at Aberystwyth University, led a project to enhance the Met Office's space weather forecasting.

Speaking from the university's AI Hub, he told me AI could offer a vital role.

"It's a very complicated system to try to model. Forecasts exist but they have many weaknesses because space weather is so complex," Dr Morgan said.

"And unlike on Earth, we can't put recording stations on the Sun or between the Sun and Earth.

"We are really dependent on remote data from telescopes.

"So AI offers a good solution, because we can't monitor the whole system constantly, and we cannot really build models that are appropriate for the system yet."

However, Dr Morgan acknowledges AI has its challenges and scientists will continue to rely on traditional space weather forecasting techniques for now.

So what about AI-generated weather presenters?

Met Office meteorologist Aidan McGivern is not so sure.

"It's important for people to have presenters they trust," he told me.

"No one wants an AI version of Derek.

"They want the real Derek - someone who can take all the data and explain it in a way that makes sense."

Aidan is optimistic about what the future holds for forecasting.

"When I started this job 18 years ago, we couldn't really predict beyond four or five days.

"Now we're giving outlooks 10 or even 14 days ahead.

"We may not be able to give specific details that far out but we can already offer a sense of whether it'll be warm or cold, wet or dry - and highlight big changes on the way.

"And with AI, the potential only grows.

"Just imagine - in the near future, we might be able to talk about a month's worth of weather at once, and visualise it in a way that really connects with the public.

"That's hugely exciting."

The potential for AI in weather prediction is immense, but AI won't replace traditional forecasting methods entirely.

It's more likely to work alongside them and be another tool for meteorologists like me to use.

For now, at least, I think my job is safe.

Hopefully the real Derek will continue to say "hello, shwmae" for a long while yet.

Additional reporting by Johnny Brew