Why are weather forecasting apps so terrible?

If you hung out laundry, visited a beach or fired up the barbecue this week, you will almost certainly have consulted a weather app first. And you might not have been entirely happy with the results. Which raises the question: why are weather apps so rubbish?

Even meteorologists like Rob Thompson at the University of Reading in the UK aren’t immune to these frustrations; he recently saw a dry night predicted and left his garden cushions out, only to find them soaked in the morning. It’s a classic example – when we complain about poor forecasts, it’s normally unexpected rain or snow we’re talking about.

Our expectations – both of the apps and the weather – are a big part of the issue here. But that’s not the only problem. The scale of weather systems, and of the data actually useful for giving us localised predictions, makes forecasting extremely complex.

Thompson admits some apps have had periods of poor performance in the UK in recent weeks. Part of the problem is the unpredictable type of downpours we get in summer, he says. Convective rain happens when the sun’s warmth heats the ground, sending a column of hot and moist air up into the atmosphere where it cools, condenses and forms an isolated shower. This is much less predictable than the vast weather fronts driven by pressure changes which tend to roll across the country at other times of year.

“Think about boiling a saucepan of water. You know roughly how long it’s going to take to boil, but what you can’t do very well is predict where every bubble will form,” says Thompson.

Similar patterns form over North America and continental Europe. But weather forecasting is necessarily a local endeavour, so let’s take the UK as a case study to examine why it’s so hard to say precisely when and where the weather will hit.

In general, Thompson is critical of the “postcode forecasts” provided by apps, where you can summon forecasts for your specific town or village. They imply a level of precision that simply isn’t possible.

“I’m in my mid-forties, and I can see absolutely no possibility during my career that we’ll be able to forecast shower clouds accurately enough to say rain will hit my village of Shinfield, but not hit Woodley three miles away,” says Thompson. These apps also claim to be able to forecast two weeks ahead, which Thompson says is ridiculously optimistic.

The two-week span was long thought to be a hard limit for forecasting, and accuracy to this day still takes a dive after that point. Some researchers are using physics models and AI to push forecasts far beyond it, out to a month and more. But the expectation we can know that much and have it apply not just globally, but also locally, is part of our disappointment with weather apps.

Despite using weather apps himself, Thompson is nostalgic for the days when we all watched television forecasts that gave us more context. Those meteorologists had the time and graphics to explain the difference between a weather front rolling over your house and bringing a 100 per cent chance of rain somewhere from 2pm to 4pm, and the possibility of scattered showers expected during that two-hour window. Those scenarios are subtly but importantly different – a weather app would simply show a 50 per cent chance of rain at 2pm and the same at 3pm in each case. That lack of nuance can cause frustration even when the underlying data is on the money.

Similarly, if you ask for the weather in Lewisham at 4pm and you’re told there will be a downpour but it doesn’t come, that looks like failure. However, wider context might reveal the front missed by a handful of miles: not failure, as such, but a forecast with a margin of error.

One thing is certain: app makers are not keen to discuss these difficulties and limitations, and prefer to preserve an illusion of infallibility. Google and Accuweather didn’t respond to New Scientist’s request for an interview, while Apple declined to speak. The Met Office also declined an interview, only issuing a statement that said, “We’re always looking to improve the forecasts on our app and exploring ways to provide additional weather information”.

The BBC also declined to speak, but said in a statement users of their weather app – of which there are more than 12 million – “appreciate the simple, clear interface”. The statement also said a huge amount of thought and user testing went into the design of the interface, adding “We are trying to balance complex information and understanding for users”.

That’s a tricky balance to strike. Even with entirely accurate data, apps simplify information to such an extent that detail will inevitably be lost. Many types of weather that can feel drastically different to experience are grouped together into one of a handful of symbols whose meaning is subjective. How much cloud cover can you have before the sun symbol should be replaced by a white cloud, for instance? Or a grey one?

“I suspect if you and I give an answer and then we ask my mum and your mum what that means, we won’t get the same answer,” says Thompson. Again, these sorts of compromises leave room for ambiguity and disappointment.

There are other problems, too. Some forecasters build in a deliberate bias whereby the app is slightly pessimistic about the chance of rain. In his research, Thompson found evidence of this “wet bias” in more than one app. He says it’s because a user told there will be rain but who is getting sun will be less frustrated than one who’s told it will be dry but is then caught in a shower. Although, as a gardener, I’m often frustrated by the inverse, too.

Meteorologist Doug Parker at the University of Leeds in the UK says there are also a wide range of apps that reduce costs by using freely available global forecast data, rather than fine-tuned models specific to the region.

Some take free data from the US government’s National Oceanic and Atmospheric Administration (NOAA) – currently being decimated by the Trump administration, which is putting accuracy of forecasts at risk, although that’s another story – and simply repackage it. This raw, global data might do well at predicting a cyclone or the movement of large weather fronts across the Atlantic, but not so well when you’re concerned about the chance of rain in Hyde Park at Monday lunchtime.

Some apps go as far as to extrapolate data that simply isn’t there, says Parker, which could be a life-and-death matter if you’re trying to gauge the likelihood of flash floods in Africa, for instance. He’s seen at least four free forecasting products of questionable utility show rainfall radar data for Kenya. “There is no rainfall radar in Kenya, so it’s a lie,” he says, adding satellite radars intermittently pass over the country but don’t give complete information, and his colleagues at the Kenya Meteorological Department have said they don’t have their own radars running. These apps are “all producing a product, and you don’t know where that product comes from. So if you see something severe on that, what do you do with it? You don’t know where it’s come from, you don’t know how reliable it is”.

On the other hand, the Met Office app will not only use a model that’s fine-tuned to get UK weather right, but it will also employs all sorts of post-processing to refine the forecasts and apply the sum total of the organisation’s human expertise to it. Then the app team goes through a painstaking process to decide how to present that in a simple format.

“Going from model data to what to present is an enormous field in the Met office. They’ve got a whole team of people that worry about that,” says Thompson. “It’s basically a subject in and of its own.”

Creating weather forecasting models, supplying them with vast amounts of real-world sensor readings and running the whole thing on a supercomputer the size of an office building is not easy. But all that work amounts to a reality we may not feel: forecasts are better than they have ever been, and are still improving. Our ability to accurately forecast weather would have been unthinkable even a few decades ago.

Much of our disappointment with the quality of weather apps comes down to demands for pinpoint accuracy to the square kilometre, to misinterpretation caused by oversimplification or to an increasingly busy public’s expectations exceeding the science.

Parker says as the capabilities of meteorologists increased over the decades, the public quickly accepted it as normal and demanded more. “Will people ever be happy?” he asks. “I think they won’t.”