US Edition
The consequence of the Great Storm of 1987 wasn’t just the loss of 15 million trees. It shook the nation’s faith in the weather forecast. After all, who can forget weatherman Michael Fish infamously declaring to viewers who were worried that a hurricane was on the way: “don’t worry, there isn’t!”
Thankfully, nearly 40 years later, scientists are a lot more confident in their data. A 15-year research programme has been launched to provide more accurate forecasts up to six weeks ahead.
The ambitious £30 million partnership between Reading University, the Met Office and the European Centre for Medium-Range Weather Forecasts will use climate data extracted from a wider variety of sources than ever before, then crunch it through supercomputers.
For anyone gambling on a sunny wedding these forecasts can’t come soon enough. But the benefits will go beyond last-minute holiday bookings, barbecues and party planning. It will transform the way industries such as agriculture, fishing and energy operate, and inform decision-making around the world to help governments protect lives and livelihoods.
“Everything we do is mainly to protect life and property and to have a thriving economy,” says Dr Florence Rabier, who is the Director General of the European Centre for Medium-Range Weather Forecasts.
“We’ve been doing these monthly forecasts for 10 or 15 years, but they’ve been improving gradually. We have more capacity with the computers now. We can go to a finer scale, making more observations. And we now have the capacity with machine learning to do better forecasts.”
Indeed, things have certainly come a long way since the Met Office was founded in 1854. It was set up by Vice-Admiral Robert Fitzroy (captain of HMS Beagle during Charles Darwin’s famous voyage) to improve understanding of marine climates and thus the safety of life and property at sea, sowing the seeds of the climate science that’s used today. The Met’s first storm-warning service followed in 1861 (this eventually became known as the shipping forecast), as did the first public weather forecast service, in what turned out to be a busy year for the fledgling organisation.
In those early years, data gathering was laborious and rudimentary. In 1877, plans were drawn up for an observatory at the top of Ben Nevis, and in 1881 the hardy meteorologist Clement Wragge climbed up to its 4,400ft summit every day during summer months to take measurements (including atmospheric pressure and wind speed and direction) before the observatory was manned year-round.
In 1922, the advancement of numerical weather predictions saw maths and physics enter the forecasting equation - effectively harnessing the study of fluids to better understand the oceans and the atmosphere. The Met Office’s first computer, acquired in 1959, turbocharged the mathematics behind this field of study, revolutionising weather forecasting forever.
A further step-change came with the availability of satellite data. In 2010 came another leap, with the use of a larger range of computers. Then, two years ago, it was machine learning. “This was when we realised we could use past forecasts and analysis alongside the knowledge we have now,” says Rabier. Crucially, weather data we have already collected is integral to our understanding of likely events in the future.
Forecasting has improved on average by a day a decade, which means in 2020 we were four days better off than in 1980, which means the accuracy of the three day forecasts at that time is now the accuracy of the seven-day forecast.
Today massive streams of data are collected from all over the globe, from deepwater buoys that warn of incoming Atlantic storms, automated weather stations that dot the countryside, weather balloons, transponders on aircraft and ships and satellites.
The new programme, which will be based at Reading University, will use even better data sources. “We’re trying to explore things that evolve on the earth a bit slower than the atmosphere,” explains Professor Pier Luigi Vidale, Professor of Climate System Science at the University of Reading and National Centre for Atmospheric Science (NCAS) senior scientist.
“Oceans give us a much better understanding of how they transport heat from the equator to the pole and influence the ways that storms develop and bring winds and rain to our shores. That will also help our forecasts.”
Another unexplored data set that the team will be drawing on is the cityscape. “Buildings and roads are not included in current climate models, yet they can have a profound influence on the weather. Current models cannot differentiate between gardens and parks or concrete and roads.” Yet heat from hot tarmac is the stuff microclimates are made of. “The stuff we use to build a city can have an impact and we need to include variables like these in our models.”
The programme will also capitalise on newly digitised older data created through an AI system that analyses weather based on historic forecasts dating back to 1940. At the moment, the AI churning runs in parallel with the more up-to-the-minute data gathering, but eventually, the two aspects will combine.
When you look at a long-range four-week forecast currently, you are looking at a weekly average. “So for instance if I am going skiing, is there a good chance there will be some snow before I go skiing? For this you could look at the monthly forecast. But you couldn’t look and say it’s going to snow a day before you arrive four weeks ahead,” explains Rabier.
Vidale lived in Colorado in the US for a long time. “There, people really respect the weather because it can be deadly,” he says. “If there is a blizzard, people are told not to go out.”
The same is true in Japan, says Vidale, where typhoons and tsunamis can have catastrophic effects. “The Japanese people really watch the apps and government warnings. Again, because it’s really deadly there.”
As the UK experiences an increase in flooding, long-range forecasting will help authorities to figure out when the rain will stop, so that they can deploy resources accordingly.
It’s also clear to see how more accurate long-range forecasts will benefit construction and agriculture industries, says Rabier: “The former needs to know when they can lay foundations of a building. In agriculture, knowing when there is a dry spell for cutting crops means you can plan ahead and employ extra workforce. Or if you can see if there’s a cold spell coming, you can decide whether to invest in covering your crops.”
And as we lean increasingly on renewable energy sources, knowing when there is likely to be more wind, sun and rain, is key for the effective deployment of wind turbines, solar panels and hydroelectricity.
But in the tourism industry, more accurate long-range forecasts could become a double-edged sword. The pros? “Climate change is going to change the tourist industry anyway but, using a long-range forecast, people can see if there’s going to be a heatwave in Spain, and if they still want to go. Or, if there’s a tropical cyclone coming in the Caribbean, then you might want to go to the Indian Ocean instead.” Which is all well and good, until you get to the cons – this could open the door to surge-pricing during periods of good weather, or consequences for holiday insurance.
But of course, this is the nation’s favourite topic of conversation. Where would we be without the ability to moan about our inclement weather? Perhaps, for all this forward-thinking, planning is overrated. Rabier says she personally doesn’t use the month-ahead forecast. “I look a few days ahead, but I don’t plan my life a month ahead. I am bound by school holidays as well. For a regular person maybe it’s not as useful as for farmers or people dealing with construction, or managing a wind farm. For me, I don’t have to plan ahead to have an umbrella.”