Ever wondered why your village was suddenly flooded by a thunderstorm the weather forecasters hadn’t mentioned? Or why they failed to warn you about the dense fog shrouding your home in the morning?
The fact is that predicting the “big picture” of future conditions has got a lot better – Storm Dennis was spotted six days before it arrived.
But getting local forecasts right – street by street and hour by hour – is still a massive challenge.
And that might now change as the Met Office secures the help of a supercomputer project costing £1.2bn.
Better forecasting means handling more data, more rapidly, and running it through simulations of the atmosphere more accurately.
Already the Met Office is pulling in more than 200 billion observations from satellites, weather stations and buoys out in the ocean every single day, and that’s set to increase.
And working out if a summer downpour will flood your home or one down the road requires more and more processing power.
“We’ll be streets ahead of anybody else,” according to Penny Endersby, chief executive of the Met Office.
“Ultimately it’ll make a difference to every individual, every government department, every industry as people see forecasts becoming steadily better.”
It’ll be the biggest investment in the 170-year history of the organisation and will dwarf the £97m bill for the current supercomputer.
In the new project, the billion-plus cost will cover not just the hardware itself but all the running costs too over a ten-year period.
There’ll be a first stage installation, which should be six times more capable than the supercomputer used now.
And then five years later there’ll be a major upgrade to increase performance by a further three times. It’ll run what the Met Office calls its “digital twin” of the Earth’s atmosphere, a highly detailed “model” of everything from the winds to the temperatures to the pressures.
To create this simulated picture of our weather, the globe is divided into grid squares.
These have become smaller as the technology has advanced – and the smaller the better because that means more accuracy.
At the moment, the model of Earth is divided up into a grid of squares that are 10km across. The UK gets more detailed treatment: its squares are 1,500m across.
London is studied with the aid of even smaller squares – 300m wide – mainly to improve the accuracy of forecasts for the airspace above the big airports.
And the ambition, when the new supercomputer is up and running, is to operate at an even sharper resolution, down to a scale of 100m.