Daily human activity largely depends on increasingly accurate weather forecasts. Millions of data points captured by satellites, balloons, ships, buoys, planes and ground stations feed complex real-time digital models that are processed on supercomputers and produce the best forecasts we’ve ever had. Without meteorological science that has overcome 200 years of obstacles and concerns, thousands of activities would be affected or paralyzed.
The graph with the evolution of the percentage of correct forecasts from the 1980s to the present leaves no doubt: a current five-day forecast is as accurate as a 24-hour forecast was in 1980 and useful forecasts (above 80 % Accuracy) are now projected up to ten days into the future. The accuracy of three-, five-, seven- and ten-day forecasts reached nearly 90% in both hemispheres and the margin of uncertainty narrowed in each.
Gone are the days when the TVE meteorologist bet on his mustache that it would rain the next day and that he would lose (as was the case for Eugenio Martín Rubio in 1967). But the mockery and reproaches about weather forecasts are still in the collective unconscious half a century later, as if nothing had changed.
These misunderstandings resurface forcefully in extreme situations like the tragic DANA in Valencia or the Filomena snowfall in 2021, despite the fact that the forecasts were the best with available technology. One cause, experts agree, is that the complexity of the models has led to the incorporation of what is called probabilistic prediction, which is very useful, but not very intuitive for the general public.
How does this system work? As the State Meteorological Agency (AEMET) explains, the European Center for Medium-Range Forecasts (ECMWF) runs 51 different forecasts on its computers, based on 51 initial states, including one undisturbed and 50 slightly perturbed. That is, we try to limit the uncertainty of a chaotic system by seeing how 50 simulations evolve in which the values are slightly different, assuming that the initial conditions are never exact.
“From these 51 predictions, the probabilities of precipitation exceeding the thresholds indicated are calculated, as well as the average and maximum precipitation expected at each point”, all with the aim of “capturing the possible developments that the atmosphere”, describes AEMET. To greatly simplify, if in 80% of these simulated scenarios it rains heavily in a region, meteorologists already have confirmation that it will be very likely.
This is the system by which the ECMWF models alerted this Sunday to the possibility that a new DANA hits eastern Spain mid-weekeven if uncertainty remains very high. The origin of this difficulty is that the equations which describe the behavior of the fluid in the atmosphere, and which are solved by supercomputers, are so complex that the search for exact solutions to the Navier-Stokes equations is one of the problems millennium mathematics for which one million euros is offered.
To understand his difficulty, the German physicist Werner Heisenberg is said to have declared that if he met God, he would ask him two questions: why relativity? and why this turbulence? “And I’m sure he’ll be able to answer me the first time,” he added.
The elusive percentages
The last time the Center for Sociological Research (CIS) surveyed the population on their confidence in AEMET, in 2011, 82.4% of those questioned considered that weather forecasts were fairly or very accurate. In this CIS survey, specific questions such as the following allow us to test our understanding of forecasts:
Suppose tomorrow’s weather forecast says “there is a 60% chance of rain.” Which of the following expressions do you think best describes what this means?
- It will rain tomorrow in 60% of the forecast location.
- It will rain tomorrow 60% of the time.
- Out of 100 days with characteristics similar to those of tomorrow, it will rain 60 days.
- 60% of meteorologists think it will rain tomorrow.
The majority of respondents (21.2%) considered option 1 correct and believed that the percentage referred to the part of the territory in which it would rain. Quite a few thought it was the proportion of meteorologists who agreed with the forecast (13.3%) and a similar percentage (12.4%) thought the key was the number of hours per day during which there would be precipitation.
The correct answer was the third (“Every 100 days with similar characteristics to tomorrow it will rain 60″ and it was the second most voted, with 19.1%. Of course, in a version updated in 2024, it would be more correct to write it in these terms: “Out of 100 forecasts run in the models, 60 predicted precipitation.“.
Uncertainty levels
“Many people wrongly think that when we talk about a 50% chance of rain, we are avoiding the problem and it’s like saying nothing,” says ETB meteorologist Arnaitz Fernández. “And others confuse the likelihood of rain with the intensity with which it will rain.” The debate is open in the scientific community itself, and many meteorologists avoid talking about percentages for fear of being misunderstood (in addition to resorting to deterministic models, which continue to have a high degree of reliability, but not both in situations like major storms or situations like DANA).
Many confuse the probability of rain with the intensity with which it will rain.
Arnaitz Fernández
— Meteorologist ETB
For Fernández, the first thing to keep in mind is that meteorology can by no means give us 100% certainty, despite improvements in satellites, observation systems and forecasts using supercomputers. “Because when something happens that the predictions didn’t predict, some people who don’t believe in science already have an argument to attack, without knowing that behind it is the atmosphere, which is a chaotic system.”
“A small variation greatly modifies the area where it will rain, and a small error in observing the initial conditions can modify the forecast,” he emphasizes. Rubén del Campospokesperson for AEMET.
What we say is what is most likely to happen, not what will happen with certainty.
Jose Miguel Vinas
— Weather forecaster
“We need to convey the idea that prediction models are already very good and very reliable, but that there is a limit and uncertainty in any prediction,” he adds. Jose Miguel VinasWeather forecaster. “But we must banish or gradually banish the idea of success and the error“. Among other things because, in the long term, it is physically impossible to be able to accurately predict the meteorological situation in a specific location, because the uncertainty is greater as the forecast area is smaller, let us remember. “What we say is what is most likely to happen, not what is certain to happen.”
Neither prophecies nor divinations
In a 2018 article on “Probability and Uncertainty in Society and Media,” meteorologists attributed public resistance to probability presentations to “the human tendency to cognitive economies » and present the truths in the simplest way possible. “A deterministic prediction (it will rain today) will be much more digestible than a probabilistic prediction (there is a 5% chance that it will rain today) and therefore, it seems to give it a divinatory character, almost beyond of scientific significance”, “they wrote “Who has never said or heard: ‘those of time have not. RIGHT Today’?”
Angel Riverastate meteorologist who worked for 38 years at AEMET, remembers the time when the means of forecasting were infinitely more precarious. “In 1982, for a situation like DANA in Valencia, what we knew was that it could rain a lot over a large part of the Mediterranean, from the mouth of the Ebro to Cabo de Gata, but we couldn’t specify more,” he explains. “I saw the beginning of numerical models, which started with resolutions of 50 km, and now we can go down to a resolution of up to one or two km.” This is why he believes that debates on the reliability of forecasts should be closed.
I experienced the beginning of digital models, which started with resolutions of 50 km, and now we can go down to a resolution of up to 1 and 2 km
Angel Rivera
— State meteorologist who worked for 38 years at AEMET
“The distrust and jokes about meteorologists go back to the days when forecasts began to be published in the mainstream media, in the early days of radio and television,” he explains. Manuel Palomaresexpert in the history of meteorology. But doubts about the reliability of predictions existed in the early days of the discipline, he recalls, when scientists themselves thought it was an incomprehensible task, due to the complexity of the atmosphere.
“Whatever the progress of science, no sincere observer concerned with his reputation will dare to predict the state of the weather,” declared the eminent French physicist and mathematician François Arago in 1846, without much success. The British Robert FitzRoy, the first to dare to publish predictions in newspapers, put his finger on the problem by inventing the word forecast (in English) in 1861. “These are neither prophecies nor predictions: the term forecast [forecast] “This strictly applies to an opinion that is the result of a combination of science and calculation,” he said. In other words, time is not guessedbut rather this was predicted from the data.
FitzRoy was vilified by his contemporaries because his predictions were not made “with precise rules” nor sufficiently supported by “observed facts”. And this harassment led him to suicide. It took nearly two centuries to develop the observation tools and machines that make it possible to predict what the atmosphere will do hour by hour all over the world and with a level of detail that seemed impossible to the pioneers. And yet, there are many who still wish lynch to meteorologists as if this scientific and technological revolution had not taken place.
