Why are weather forecast applications so terrible?

If you have dragged the laundry, visited a beach or put on the barbecue this week, you will have almost certainly consulted a meteorological application first. And you may not be completely satisfied with the results. Which raises the question: why are weather applications so waste?

Even meteorologists like Rob Thompson at the University of Reading in the United Kingdom are not immune to these frustrations; He recently saw a predicted dry night and left his garden cushions, only to find them soaked in the morning. This is a classic example – when we complain about bad forecasts, it is normally an unexpected rain or snow that we are talking about.

Our expectations – applications and weather – are a large part of the problem here. But this is not the only problem. The scale of meteorological systems, and data really useful to give us localized predictions, makes forecasts extremely complex.

Thompson admits that some applications have had periods of poor performance in the United Kingdom in recent weeks. Part of the problem is the unpredictable type of showers we get in summer, he says. The convective rain occurs when the heat of the sun heats the floor, sending a column of hot and humid air in the atmosphere where it cools, condenses and forms an isolated shower. It is much less predictable than the vast meteorological fronts driven by pressure changes that tend to roll across the country at other times of the year.

“Remember to boil a saucepan of water. You know roughly how long it will take to boil, but what you cannot do very well is to predict where each bubble will form, “explains Thompson.

Similar models are formed on North America and continental Europe. But weather forecasts are necessarily a local business, so let’s take the United Kingdom as a case study to examine why it is so difficult to say precisely when and where time will strike.

In general, Thompson criticizes the “postal code forecasts” provided by applications, where you can summon forecasts for your specific city or village. They involve a level of precision which is simply not possible.

“I am in mid-quanta, and I see absolutely no possibility during my career that we will be able to plan the shower clouds quite precisely to say that the rain will strike my village of Shinfield, but not hit Woodley at three kilometers,” explains Thompson. These applications also claim to be able to predict two weeks in advance, which, according to Thompson, is ridiculously optimistic.

The duration of two weeks was considered a difficult limit for forecasts, and the precision to date always takes a dive after this point. Some researchers use physics models and an AI to push forecasts far beyond, a month and more. But the expectation that we can know a lot and that it applies not only in the world, but also locally, is part of our disappointment with weather applications.

Despite the use of weather applications itself, Thompson is nostalgic for the days when we all watched the television forecasts that gave us more context. These meteorologists had time and the graphics to explain the difference between a meteorological front riding your home and bringing 100% of rain somewhere from 2 p.m. to 4 p.m., and the possibility of dispersed showers expected during this two -hour window. These scenarios are subtly but above all different – a meteorological application would simply show a chance of rain of 50% at 2 p.m. and the same at 3 p.m. in each case. This lack of nuances can cause frustration even when the underlying data is on money.

Likewise, if you ask Lewisham the weather at 4 p.m. and you are told that there will be a downpour but it does not come, it looks like a failure. However, a broader context could reveal the front failed by a handful of kilometers: not failure, as such, but a forecast with a margin of error.

One thing is certain: application manufacturers are not eager to discuss these difficulties and these limits, and prefer to preserve an illusion of infallibility. Google and Accuweather did not respond to New scientist‘s Request for an interview, while Apple refused to speak. The Met Office also refused an interview, published only a declaration which said: “We are still trying to improve forecasts on our application and to explore means to provide additional weather information”.

The BBC also refused to speak, but said in a statement users of their meteorological application – including more than 12 million – “appreciate the simple and clear interface”. The declaration also indicates that a huge reflection and user tests have been devoted to the design of the interface, adding “we are trying to balance complex information and understanding for users”.

It is a delicate balance to find. Even with fully precise data, applications simplify information to such an extent that the details will inevitably be lost. Many types of time that may be radically different to experiment are grouped in one of the rare symbols whose meaning is subjective. How many cloud cover can you have before the sun symbol is replaced by a white cloud, for example? Or a gray?

“I suspect that if you and I give an answer, then we ask my mom and your mom what it means, we will not have the same answer,” explains Thompson. Again, these types of compromise leave room for ambiguity and disappointment.

There are also other problems. Some forecastists build a deliberate bias by which the application is slightly pessimistic about the chances of rain. In his research, Thompson found proof of this “wet bias” in more than one application. He says that it is because a user said that there would be rain but who will become the sun will be less frustrated than the one who said that he would be dry but that he is then caught in a shower. Although, as a gardener, I am often frustrated by the opposite too.

Meteorologist Doug Parker at the University of Leeds in the United Kingdom says that there is also a wide range of applications that reduce costs using global forecast data available for free, rather than refined models specific to the region.

Some take free data from the National Oceanic and Atmospheric Administration of the US Government (NOAA) – currently decimated by the Trump administration, which puts the precision of forecasts in danger, although it is another story – and simply reconditions. These raw global data may well foresee a cyclone or the movement of large meteorological fronts across the Atlantic, but not so well when you are concerned with the risk of rain at Hyde Park at lunchtime on Monday.

Some applications go so far as to extrapolate data that is simply not there, explains Parker, which could be a question of life and death if you try to assess the probability of sudden floods in Africa, for example. He saw at least four free forecasting products of a dubious utility has radar data for precipitation for Kenya. “There is no precipitation radar in Kenya, so it is a lie,” he said, adding that satellite radars are going intermittently to the country but do not give complete information, and his colleagues from the Kenya Meteorological Department said they did not have their own radars. These applications are “all produce a product, and you do not know where this product comes from. So, if you see something serious about it, what are you doing with it? You don’t know where it comes from, you don’t know how reliable it is. ”

On the other hand, the Application Met Office will not only use a refined model to obtain the right British weather, but it will also use all kinds of post-processing to refine forecasts and apply the total sum of human expertise of the organization. Then, the application team goes through a meticulous process to decide how to present it in a simple format.

“Speaking from the data of the model to what to present is a huge area in the Met Office. They have a whole team of people who are worried about it, ”explains Thompson. “It is essentially a subject in itself.”

Creating weather forecasting models, providing them with large amounts of real -world sensor readings and managing everything on a supercomputer the size of an office building is not easy. But all this work is equivalent to a reality that we may not feel: the forecasts are better than they have ever been and always improve. Our ability to predict the weather precisely would have been unthinkable a few decades ago.

A large part of our disappointment concerning the quality of weather applications comes down to requests for punctual precision per square kilometer, a misinterpretation caused by excessive simplification or the expectations of the increasingly occupied public exceeding science.

Parker says that the ability of meteorologists have increased over the decades, the public quickly accepted it as normal and asked more. “Will people never be happy?” he asks. “I think they won’t do it.”