Professional Meteorologists
You’ve heard it before — at cookouts and in group chats and in line at the store when the sky is doing something ambiguous. They're never right. It’s said with satisfaction, like the way complaints about the DMV are said. Nobody ever asks for evidence, because everyone in earshot has their own — the ruined weekend, or the eleven inches of snow that turned out to be two. The claim is treated as folk knowledge.
The person delivering the line checked the forecast that morning. They'll check it again. They have an app on their phone that they have opened, conservatively, ten thousand times. If you took it away they would be genuinely inconvenienced within a day.
The complaint is aimed at what is one of the most successful sciences of the last fifty years. The Met Office puts it plainly: its four-day forecast is now as accurate as its one-day forecast was thirty years ago.1 In the 1970s, the National Hurricane Center's three-day track forecasts — the prediction of where a hurricane will actually arrive — were off by more than four hundred nautical miles on average. Over the last five years the same forecast has averaged an error of eighty-six.2 Better observations, faster models, better ways of getting the observations into the models, compounding for decades, until predicting the future became an ordinary municipal service that appears on your lock screen for free.
So I went looking for the distrust, expecting the numbers to be grim. They aren't. In April 2023, YouGov asked Americans how much they trusted fifty-six news organizations — broadcast, print, digital, social. The most trusted outlet in the country was The Weather Channel, and it was the only one on the entire list that both Democrats and Republicans trusted more than they distrusted.3 A separate YouGov survey that same spring put the National Weather Service at 66% trustworthy and NOAA at 59%, with only five or six percent calling either one untrustworthy.4 Against the current baseline for public confidence in institutions, those are extraordinary numbers.
Which means the folk knowledge is wrong. People do trust meteorologists. They trust them more than almost anyone else who talks to them through a screen. They just also say, constantly and with feeling, that meteorologists are always wrong.
Part of the gap is due to human memory, which is not a reliable instrument. The forecast is right on Tuesday and right on Wednesday and right on the nine days after that, and none of those days produce a story. There is no version of the anecdote that begins so I checked the weather, and then that's exactly what happened. The correct forecast disappears into the day it correctly described. The bust, meanwhile, has a plot: expectation, betrayal, consequence. It gets told at dinner. It gets told for years. Two hundred confirmations and one ruined Saturday do not weigh the same, and no amount of accuracy fixes that.
The tempting second explanation is that people can't read a probability, such as Forty percent chance of rain. But when YouGov showed Americans a forecast reading "Precipitation: 80%" and offered four interpretations, 63% picked the right one. The most popular wrong answer — that 80% of the area would get rain — drew 15%.5 That's not a public failing to understand; that's a solid majority understanding a technical sentence correctly. And the age breakdown runs backwards from what you'd guess: 88% of people 65 and older got it right, against 41% of adults under 30. The group raised on weather apps understands the number in the app the least.
What the forecast is actually reporting is called the probability of precipitation, and the National Weather Service defines it: the chance that one specific point in the forecast area gets at least a hundredth of an inch.6 The professionals don't agree on it either. When researchers surveyed 188 meteorologists and broadcasters for a 2016 paper in Weather and Forecasting, they got a range of different definitions, each offered confidently — and 43% of the meteorologists said there was little or no consistency in how the term is defined at all.7 The public is handling this number about as well as the people who publish it.
So innumeracy isn't the answer either, and something underneath the definitions stays unfixed regardless. A probability is a claim about a distribution, and you get to live exactly one outcome. When the forecast says forty percent and the day stays dry, nothing has been disproved and nothing has been confirmed; you have drawn one card. To actually evaluate that forecast you'd need the other ninety-nine Saturdays, and you don't have them — you have the one, and it had your kid's game in it. In aggregate the forecast is rigorously checkable. In a life, it is not.
And yet: it feels checkable.
Consider how little of the expertise in an ordinary life ever gets graded where you can see it. Your cardiologist makes a judgment about your next ten years and neither of you will ever isolate whether it was right. An economist explains what interest rates will do, and by the time the answer arrives the question has been reworded. City planners, actuaries, the people who decided how your road drains — all of them are making predictions, and essentially none of them come back the next day to be scored. Most expertise is protected by time, complexity, and the sheer difficulty of ever running the comparison.
The meteorologist has none of that protection. They make a specific, falsifiable claim about tomorrow, in your actual city, in public, under their own name — and tomorrow arrives on schedule, for everybody, at the same time, and everyone has a window. They have to come back and do it again in front of the people who just checked their work. There are few other professions that submits to that, daily, at that volume, in front of that many amateur auditors.
So that complaint may not be a symptom of distrust at all. It may be what accountability sounds like from the outside.
There's another sentence, delivered in the same tone by the same people, that has nothing to do with accuracy. It’s said when the weather is simply bad. Not wrong — bad. The forecast said rain, and it rained, and the day was ruined as promised, and somebody is still furious at the man on television. I assumed for years that this was a bit, like booing the umpire. I’m sure some of it is. But not all of it.
We say they're calling for rain, which is a summoning verb — nobody says the astronomer is calling for an eclipse.
Psychologists have caught people doing this under laboratory conditions. In a 2019 paper in the Journal of Experimental Psychology: General, Leslie John, Hayley Blunden, and Heidi Liu ran eleven experiments on what they called, plainly, shooting the messenger. In one, participants stood to win fifty cents by guessing whether the next day's Wall Street Journal headline would have an odd or even number of words, and a research assistant told them how it came out. The ones who lost rated the assistant less likable — and reported believing she had been hoping they would lose. She had not written the headline, counted the words, or stood to gain anything either way. The effect comes from a desire to make sense of situations that don't make sense, and from a tendency to assign bad motives to whoever is standing there when they happen. It gets stronger the less sense the bad news makes. Told their flight was delayed three hours, people blamed the airline employee more when another plane had taken their departure slot than when the delay was simply the delay — a scenario the authors describe as violating "the commonly held beliefs that the world is just, predictable, and comprehensible."8 Randomness is harder to sit with than malice.
James Spann has forecast weather in Birmingham for more than forty years. In January 2019 he called for one to two inches of snow across north and central Alabama. He published the uncertainty as an explicit range — a "boom" value of four inches, a "bust" value of a quarter inch — and wrote that some people would get nothing at all. The snow arrived on time. It didn't accumulate; the ground was too warm, and the roads stayed clear. What followed was threats, vulgar profanity, and "the type of vitriol you never want to hear."9 Nobody had been endangered. Nobody had been sent unwarned into an ice storm. What had happened is that a lot of children didn't get a snow day, and the disappointment needed somewhere to go.
Historians of witchcraft have documented weather magic as one of the most common accusations in early modern Europe — witches held responsible for thunder, lightning, hail, snow, and frost — and the worst outbreaks cluster in the coldest stretches of the Little Ice Age, when crops failed and famine followed. In 1654 a woman named Katherine Grady was hanged aboard a ship bound for Virginia because the passengers concluded she had conjured the storm that was nearly sinking it. Emerson Baker, a historian of the Salem trials, notes that Salem itself fell in the worst weather of the entire Little Ice Age. He put it simply: the more miserable people are, the more likely they are to look for someone to blame. He concludes by observing that we no longer accuse anyone of witchcraft — that these days it is the meteorologists who catch it.10
Bad weather didn't produce witch hunts. What it produced was misery, in a culture that predetermined who was responsible for misery. That determination never involved explaining the storm. It involved finding the person behind it — and when a culture offers up a candidate, the candidate gets blamed. Ours offers one who volunteered: someone who goes on television, gives a name, and agrees to be there again tomorrow.
Which is why the two complaints only sound alike. One is an audit, and it holds up: heckling the forecast is what it sounds like when a prediction can actually be checked. The other was never a claim about the forecast at all, so no improvement in forecasting will touch it. It's grief that has found a direction. Spann's forecast was defensible and the anger came anyway.
The forecast is drifting away from the person delivering it. Weather apps are now the most common source, used by 53% of Americans against 43% for TV weather stations. The ritual is thinning too: 85% of Americans 65 and older check a forecast at least daily, against a third of adults under 30.11 There may be nothing to the fact that the youngest, most app-native cohort is also the one that reads the percentage least accurately. But what's being removed is the arrangement where a person tells you something and then has to stand there afterward. When the last local meteorologist is gone, the forecast will still be right, and there will be nobody to accuse.
There is a political angle too, since weather sits next to climate. Chris Gloninger was chief meteorologist at the CBS station in Des Moines when he started drawing the line between the storms he was covering and the climate producing them. In June 2023 he left television, citing a death threat over that coverage and the PTSD that followed. The man who sent the emails pleaded guilty to harassment.12
But Gloninger insisted on something afterward. It was one person. He asked that Iowans not be judged by it, said the overwhelming majority of people he'd met there were kind, and printed out the hundreds of supportive emails he got after announcing he was leaving. A single obsessive man is not a polarized electorate. The polling says the same thing from the other end: an institution that has been successfully politicized does not come out of a national survey as the one thing Democrats and Republicans both say they trust. Something real is happening at the seam between weather and climate, and it is costing people. It still doesn't explain the distrust.
The complaint may actually be a sign of health. An institution checked by millions of amateurs every day, holding up, and grumbling about anyway is in better shape than one that never has to be right in public. The grumbling is the receipt.
The grief is not. It isn't about the meteorologist, and they catch it regardless, and no improvement in the models will touch it — because the impulse behind it is older than radar, older than the profession, older than the idea that the sky follows rules at all.
We aim our skepticism at an institution that has earned the least of it. The fact: when something goes wrong and nobody did it, the anger doesn't evaporate. It goes looking. It always has.
Sources
1. Met Office Press Office, "Why an 'exact date' weather forecast headline isn't what it seems" ↗, 15 November 2023: "the Met Office's 4-day forecast is as accurate as its 1-day forecast was 30 years ago." ↩
2. Current figure from the National Hurricane Center, Official NHC 5-Year Average Forecast Errors, 2021–2025 ↗: Atlantic basin mean official track error at 72 hours is 86.4 n mi across 805 forecasts. Historical figures from NHC's official error trends ↗ (Atlantic track errors by decade, 1970–2025), as charted by Our World in Data, "Weather forecasts have become much more accurate" ↗, 12 March 2024. Note that Our World in Data renders the modern value as "less than 80 miles"; the NHC's own five-year table gives 86.4 nautical miles, and the figure in the text follows the NHC. ↩
3. YouGov, "Trust in Media 2023: What news outlets do Americans trust most for information?" ↗, 8 May 2023; poll of 1,500 U.S. adult citizens, 3–9 April 2023, margin of error ~3%, covering 56 outlets. The Weather Channel led on net trust (+53) and was the only outlet with positive net trust among both Democrats (+64) and Republicans (+47). ↩
4. YouGov, "How often and where Americans get information on the weather" ↗, 8 May 2023; poll of 1,000 U.S. adult citizens, 14–20 April 2023, margin of error ~3%. NWS 66% and NOAA 59% "very or somewhat trustworthy"; 6% and 5% untrustworthy respectively. ↩
5. Same poll as note 4. Shown "Precipitation: 80%" with four interpretations, 63% chose "There is an 80% probability that it will rain tomorrow in your area"; 15% chose "80% of your area will receive rainfall tomorrow"; 88% of respondents 65 and older answered correctly against 41% of adults under 30. ↩
6. National Weather Service, "What is the Meaning of PoP" ↗: the probability of precipitation "describes the probability that the forecast grid/point in question will receive at least 0.01″ of rain." ↩
7. Stewart, A. E., C. A. Williams, M. D. Phan, A. L. Horst, E. D. Knox, and J. A. Knox, "Through the Eyes of the Experts: Meteorologists' Perceptions of the Probability of Precipitation" ↗, Weather and Forecasting 31(1), 2016, 5–17, doi:10.1175/WAF-D-15-0058.1. Survey of 188 meteorologists and broadcasters; varied definitions held with high confidence; 43% reported no or very little consistency in how PoP is defined. ↩
8. John, L. K., H. Blunden, and H. Liu, "Shooting the Messenger," Journal of Experimental Psychology: General, 2019, doi:10.1037/xge0000586; eleven experiments. The Wall Street Journal headline study and the malevolent-motive finding are described in the British Psychological Society's Research Digest summary ↗, 8 May 2019. ↩
9. James Spann, "Snow Forecasts, Communicating Uncertainty, and Social Media Hissy Fits" ↗, 1 February 2019. Contains his 28 January forecast discussion verbatim, including the boom and bust values and "some will get nothing at all," and his account of the response. ↩
10. Emerson W. Baker, "Winter, weather, and witchcraft" ↗, OUPblog, March 2015, drawing on Wolfgang Behringer, Witches and Witch-Hunts. Source for the Katherine Grady hanging (1654), the Little Ice Age correlation, the Maunder Minimum coinciding with Salem, the scapegoat mechanism, and the closing line about meteorologists. ↩
11. Same poll as note 4: weather apps the most common source at 53%, TV weather stations 43%, weather websites 34%, local TV news 27%; 60% of Americans check a forecast at least daily, rising to 85% of those 65 and older and falling to 33% of adults under 30. ↩
12. Rachel Treisman, "A meteorologist got threats for his climate coverage. His new job is about solutions" ↗, NPR, 27 June 2023. Gloninger was chief meteorologist at KCCI-TV in Des Moines; he announced his departure citing "a death threat stemming from my climate coverage last year and resulting PTSD, in addition to family health issues." The sender was identified as Danny Hancock, who pleaded guilty to third-degree harassment. Gloninger's own caveats — that it was one individual, that he didn't want Iowans judged by it, and that he received hundreds of supportive messages — are from the same piece. ↩