The real reason weather forecasters (like me) often appear to get it wrong

Sometimes I’m going to walk around in a supermarket, and a shopping will approach me in the hallway. They said, “I organized a barbecue on Saturday and you told me it would rain.” “And he didn’t do it. Why did you get it wrong?”

Or the opposite: One day they planned the sunlight, only disappointed by the gray sky. Or a parent can ask me how the weather could be for the wedding of his sons in March – in September.

These people are always an exquisite friend, and speeches are part of what I have made the air I have done for the last thirty years.

But they also shed light on a strange truth.

Throughout my career, guessing has gone beyond recognition. Now we can guess with higher accuracy and more detailed details than I started to offer the air in the mid -1990s.

Liz Bentley, Professor of Meteorology at Reading University and General Manager of the Royal Meteorology Association, says that one -day prediction is more than 90% of the time.

However, despite these steps, there are still gaps in the trust of the people.

When Yougov asked British adults last summer, whether they trust the weather forecast, a significant minority – 37% – 37% – they said they didn’t trust “too much” or “at all”. (Reassuring, 61% said that they trusted to predicts like me.)

Jokes about estimation are common. The opening ceremony of the 2012 Olympics, the air predictive Michael Fish told the audience not to worry about a moment of a moment, because only one storm to hit hours later.

(As it is, Michael was true: hurricane winds hit Southeast England that night, but technically not a hurricane.) Nevertheless, the event became a word for an estimated error.

So, with our wealth of knowledge and strong estimation technology, some people still perceive the air as wrong? And do we really understand it wrong, or is it more complicated in the game around how we share the estimates?

Great accuracy and great expectations

Part of the difficulty is around the expectations rising in our world of access to information day by day.

We can change the temperature of our refrigerator or define a problem in our car in some of our smartphones. So why can we not find that it will rain on the street with 100% accuracy on Sunday at 14:00 – of course an easier success?

Another part of the difficulty is how this wealth of knowledge is boiled and transmitted.

Meteorology produces overwhelming amounts of data; Snappy, TV or digital application friendly to focus on an estimation. This means that even if we are technically correct, some viewers can still be confused.

However, the answer also lies in the difficult nature of meteorology.

This is a sensitive science and any small mistakes in the data may distort or remove things from the shape.

Every day, throughout the British Islands, predictions, more than 200 “air station” network operated by Met Office, collect “observations” (or data) on things such as temperature and wind speed. The data are then attached to mathematical models run by powerful machines or “super computers”.

Earlier this year, Met Office introduced a new super computer from a physical machine to cloud -based software for the first time.

He says that the new device will “offer better forecasts and help scientists advance important climatic research worldwide.”

However, as in any science, there are weaknesses.

Chaos theory: when the weather goes upside down

The atmosphere is known as a “chaotic system”, ie a slight error, even small as 0.01c in the first observations, can produce a great extent different result.

Bent This is called chaos theory, Prof Bentley explains. “Or butterfly effect. Analogy, if a butterfly beats its wings in Brazil, it may have an effect on the atmosphere in Northern Europe after six days.”

In addition, there is a special difficulty in predicting air in small geographical areas.

In the 1990s, an air incident had to be larger than about 100 miles (161km) before it was fully observed – now, the UK -diameter air model used by Met Office could map the weather events in a small way, such as 2 miles (3km).

However, it is difficult to get closer beyond this dimension, so it is particularly difficult to predict the air as heavy fog that can affect an area of only 1 km.

And even with great developments in science, technology disruptions still occur – but these are rare with compassion.

In the autumn, the BBC Weather website briefly showed over 13,000Mph in London and showed 404C temperatures in Nottingham.

The BBC apologized to “a problem with some weather data from our forecast provider”.

Problem with boiling data

The biggest challenge of my job is to synthesize this data to fit into a strict television segment.

“There is no other science that is tested, controlled and tried by the general public,” he says.

“Nuclear fusion physics are as complex, but most of us do not experience it from that day to day, so we don’t have to find a way to convey this science to the public.”

It is also easy to forget that this is just this.

Over the years, we have been much better in this subtle “transmission of uncertainty” art. Meteorologists now produce “community forecasts”, where they can run 50 different models, all of them have small variations.

If all these scenarios point to a similar result, meteorologists can be sure that they do it correctly. If they produce different results, their trust is much lower.

Therefore, in an weather application, you can see a 10% chance of rain in your area.

Time to rethink predictions?

Estimators often think of this difficult communication issue; How the air can be explained more easily.

Last week, BBC is a New partnership with Met Office. He came eight years after he officially ended his relationships (since 2018, the Netherlands made meteogroup BBC estimates).

The new agreement aims to combine the expertise of the two organizations and to “transform science into stories”.

Of course, some think that more creativity is needed to transmit weather. Dr Hosking from Alan Turing Institute argues that predictions can move away from giving rain chances, and instead they can use the “story approach”.

In this way, predictions can say things like “what we see right now is like what we see in a certain event a few years ago” – something in memory. “

Therefore, in 2015, the Met Office decided to name storms.

However, Prof Bentley argues that numbers can be strong and perhaps it is better to arrange consumers with difficult data they need.

He says that air estimation in the United States has percentages “everywhere”; American consumers are talking about everything from the chance of rain to the spread of temperature.

“People are comfortable [with it]”he says.

New Air Super Figctor

Air estimation may change significantly with the emergence of artificial intelligence (AI). The use of machine education to estimate the air has developed rapidly in recent months.

Generally, estimators are said to have gained 24 hours of accuracy in each other in decade, ie Met Office can now publish an air warning seven days in advance.

However, the AI models designed by Google Deepmind are estimating 15 days in advance, Dr Hosking says.

Earlier this year, a researcher team from Cambridge University published a completely AI -guided air program called Aardvark Weather. The results were written in Nature Journal.

While traditional estimate requires hours of use on a strong super computer, researchers say Aardvark can be deployed on a desktop computer in a few minutes. They claim that it uses less information processing for “thousands of times” and can predict the air with more details.

They also claim that they will improve forecasts in West Africa and other poor regions (the best traditional forecast models are mostly designed for Europe and the US).

“It can be a transformation; it can be a transformation,” he says.

However, Prof Bentley defines a weakness in AI -guided air models: they are fed by scratching historical data and are trained to detect patterns – this makes it very difficult to predict events that have not yet occurred.

“We will see new records with climate change, or he says. “We can see 41c in the United Kingdom. However, if AI is always looking back, he will never see 41 because we haven’t had it yet.”

Prof Turner acknowledges that this is a challenge in AI models like him and says his team is working on drugs.

‘So what’ factor

In the future, analysts think that predictions will be more depth. Instead of just guessing the rain, guessing what the rain will have an effect on you – will say more and more and more in your garden plans.

Prof Bentley calls it “what” factor. “Are you putting something on it [a weather app] This said, ‘If you are planning a barbecue, you may want to make lunch, because do you have a chance to wash in the afternoon?’

This is bell with a trend that I noticed from my own career: the increasing interest in understanding science behind the air.

The viewers are no longer interested in knowing if there is only one heat wave; They want to know why.

This is why we publish more content that explains the physique of Aurora Borealis, or why climate change leads to larger stones.

As for artificial intelligence, it can definitely increase accuracy – but there is a risk of greeting with information. Dr Hosking says that artificial intelligence is more agile and can change the air models faster, and users can soon reach frequently changing estimates. They can also have “much more localized” information (perhaps not only in your city, but also in your backyard, other analysts estimate).

This can lead to an overwhelming amount of data for those who use the application and stick users to their smartphones. And in this world, it will be even more important for human predictions to convey the air in a clear and understandable way.

But there are also fronts – at least much longer term, not expecting more accurate predictions.

Maybe one day, when a mother asks me to guess at the wedding at the wedding six months after me, I can give a little better answer.

Additional Reporting: Luke Mintz

More than Devred

BBC Independent With the new perspectives that challenge assumptions and make deep reporting about the biggest problems of the day for the home and the best analysis on the website. And we exhibit thought -provoking content from all over BBC Sounds and Iplayer. You can send us your feedback in the Dextth section by clicking the button below.