Is Your Weather App Awful? It Might Be Your Fault.
Slate · L · trust 54/100
Follow Nitish Signed Up For Email Alerts Error Signing Up For Email Alerts Close Enter your email to receive alerts for this author.
Sign in or create an account to better manage your email preferences.
Are you sure you want to unsubscribe from email alerts for Nitish Pahwa ?
A few weeks ago, I biked to meet a source for coffee, since my Android phone’s AccuWeather widget had seemingly projected a low, low chance of precipitation in the area. You can guess what happened then. The clouds decided to cluster above and surprise me midtrip, releasing the rain down hard as I finished the journey, weaving as strategically as possible to avoid getting splashed by the cars next to me. I showed up to the appointment late, soaked, and embarrassed.
As it turns out, I’m hardly the only person to have felt hoodwinked, bamboozled, and/or flat-out deceived by my smartphone’s forecasts in 2026. Throughout this year, the entire country seems to have been afflicted with the weather-app blues. Over in San Francisco, which faced down a severe heat wave in March, Google Weather users were left flummoxed by the charts of low-70s temps , way milder than the 90-degree record felt by downtown residents. Not long after, Philadelphians underwent a round of spring-showers confusion on April 22, when the Weather Channel told them that the storms would quiet down at 7:15 a.m. … even though it hadn’t rained at all that morning.
“I honestly think weather apps have gotten worse in the new AI world we live in,” a North Carolina meteorologist wrote on Facebook in December, noting that Apple’s weather app got the forecast that day “completely wrong.” Their conclusion: “Human-made forecasting is still the best way to go, and even that is still wrong.”
Such incidents have left many Americans mistrustful of weather apps as a whole, and desperate for viable alternatives. The stakes go far beyond daily routines or vacation itineraries; small businesses lose out on potential foot traffic and revenue whenever a thunderstorm icon appears, even if it’s bright and sunny outside. Yet it’s difficult to know where to turn. Social media platforms and message boards are crawling with asks for reliable weather app recommendations, and the answers can vary wildly. Vincenza Berardo, a philosophy instructor based in Massachusetts, told me she has four alternate weather apps on her iPhone— EverythingWeather , MyRadar , WTForecast , and Weather Underground —but finds them all equally likely to forecast rain on days when it doesn’t, and vice versa.
Complaints over smartphone weather apps are nothing new; my colleague Lizzie O’Leary has been on this beat for years. But they feel especially discombobulating and urgent at a time when the United States’ world-class weather-monitoring infrastructure has been decimated , private actors are attempting to fill the monitoring gaps, and the changing climate makes freak events like this summer’s heat domes, floods, and wildfires harder to prepare for.
Maybe, however, it isn’t all on the apps themselves. “Users have always thought that weather apps are inaccurate, but they’re also incredibly untrustworthy about their experiences,” Steve Gifford, co-founder and CEO of the data-processing startup Wet Dog Weather , told me.
Daniel Swain, a California-based climate scientist, tends to agree with that assessment. “People’s expectations are increasingly misaligned with what the apps are actually telling you, and with what weather prediction can tell you in the first place,” he told me.
As Swain described it, most weather apps tend to be “opaque” in disclosing what informs their forecasts, but there are a few key, gold-standard data sources utilized by almost every relevant program on the app store. “Global weather forecasts are made using predictive models—huge mathematical systems with thousands of lines of code, run on supercomputers,” he said. “Every part of the Earth’s surface is mapped out into a grid cell with relevant information, and that’s run through the models to create initial predictions that get assimilated, often in raw form, by app data providers.”
The computational and informational resources involved in these calculations are hefty. As such, the most thorough, wide-spanning modeling sources belong to governmental organizations that can afford the setups, including the Global Forecast System (run by the U.S. National Weather Service) and the European Centre for Medium-Range Weather Forecasting, among others. The best-known and biggest brands in forecasting—Apple Weather, AccuWeather, Weather Underground, Weathergraph—rely on both, but tend to especially pull from GFS, which is run four times a day and estimates what conditions may look like up to 16 days in advance. “It all tends to be the same data,” said Gifford. “If some app tries to tell you it has a fancier forecast, you roll your eyes a bit, because it all mostly comes from governments.”
It’s true that the NWS’s modeling capacity has been constricted thanks to the DOGE-ordered budget-slashing that hit the National Oceanic and Atmospheric Administration last year , which forced federal scientists to cut back on regional office spaces as well as satellite and balloon launches that help them track weather patterns. (Gifford, for his part, thinks the European Centre has a better modeling system, since it refreshes its daily calculations far more often than GFS does.) And yes, Swain told me, many predictive data sources that relied on physics-based modeling now utilize a “ blend ” of physics-based and machine learning/artificial intelligence methods. But those core modeling differences, which fuel what Swain calls “slight but significant differences” in projecting baseline conditions, don’t make much of a difference as to what reaches your app—because those displays are determined by how appmakers and developers actually curate and present the data at hand.
“Apps have access to various possibilities predicted by various models—but…
Read the original at Slate →
Open in TruthVane →