:format(webp)/cdn.vox-cdn.com/uploads/chorus_image/image/72791001/1638763606.0.jpg)
For the last seven weekends in New York City, it has rained. Never during the week. Never for just an hour. Never not inconvenient. At this point it feels like the inclement weather has gone fully sentient, and knows the exact time on Friday to start ruining New Yorkers’ plans.
Over this time, this relentless weekend-only rain has also affirmed that Apple’s weather app is pretty much useless. Personally, I’ve learned that the app cannot distinguish between “light rain” and “rain,” that the percentages it spits out feel bogus, and to never trust it when it tells you what time the rain will stop. I’m not alone. My friends and coworkers also have various stories about how the app has let them down, or how sometimes it just won’t work. Some even talk about Dark Sky, a weather-forecasting app that Apple bought in 2020, with a mournful, wistful sadness, like a lost love. Apple says Dark Sky’s most beloved features have been integrated into its app, but Dark Sky fans aren’t convinced. Things were different then, they say. Things were better.
My growing frustration spurred me to find out why Apple’s weather app stinks. In speaking to experts, I was comforted by the fact that there’s actually a reason — algorithms, specifically — for my annoyance. It’s nice to be mad at something in particular. But in my search I also discovered newfound appreciation for local meteorologists and more about weather and weather forecasting than I had initially planned.
What we mean when we say Apple’s weather app stinks
My serious complaint with Apple’s weather app is that it won’t give me a straight answer when it comes to rain. Rain means wet socks, puddles, a dampness in my clothes that hangs around all day. It also means dealing with people who say, “Oh, we needed this” with a polite smile.
Rain sucks.
My needs are simple: I want to know if it’s going to rain, how much it’s going to rain, when the rain will start and when it’ll stop. Ideally, I would like to not have to go outside to check if it’s raining, because why else would I have a powerful computer in my hand if it couldn’t tell me things that were happening around me?
:format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/25025774/1544682674.jpg)
“The Apple weather app is not good for specifics,” says John Homenuk, the meteorologist behind NY Metro Weather. Homenuk has gained a loyal New York City following for his accurate and jaunty daily weather forecasts. “And, unfortunately, specifics is what we need if we’re planning our life. ‘Do I need a jacket tonight? Is it gonna rain when I go to sit on the rooftop later?’ It struggles with that type of stuff.”
Homenuk explained to me that Apple’s weather app, and weather apps in general, work by using algorithms to interpret data — weather models, location, current observations — culled from various sources. Other experts I spoke to said apps don’t disclose what data they’re using nor how frequently they source the data, which can lead to imprecise readings.
These algorithms also have limits. In weather forecasting, these limits show up because those equations are based on models that meteorologists understand to be imperfect.
“There’s one big model that is used not only in apps, but weather data around the United States. It’s called the GFS, the Global Forecast System,” Homenuk said, adding that the GFS tends to err on the side of speed, sometimes projecting storms going out to sea and out of the area faster than anticipated. Meteorologists who understand the GFS know its faults, and use those faults and what the GFS is predicting to provide a more accurate forecast.
“If there’s a snowstorm developing ... the app could be showing that four days from now it’s going to be sunny and 45 degrees because the app’s using the GFS. But we know as human beings that this model always does this. It’s always too far out to sea with the storms, and we’ll be more careful,” Homenuk said, providing a hypothetical example.
The GFS is just one model of many, and each one has its own tendencies and errors that humans can correct for. Algorithms don’t have that kind of discernment yet, which in turn makes app predictions like precipitation and storms somewhat imprecise. Algorithms also can’t compete with the human experience of living somewhere and knowing how weather behaves in that particular area.
“Terrain can have a huge effect on how those models perform,” Jeff Givens, a meteorologist based in Durango, Colorado, told me over email. Givens’s accurate forecasts (especially when it comes to snow and storms) have garnered him a following on his extremely popular site Durango Weather Guy, because the San Juan Mountains tend to bork general weather predictions in his area. “Apps and models perform better in flat terrain.”
Given this information, it seems like weather apps perform best in places with predictable precipitation patterns, as well as places where there aren’t mountains or any kind of topographical features to skew things. People in Southern California probably do not complain about Apple’s app as much as someone in Durango or even New York City would.
Apps are great, when you put them into perspective
When I asked Alexander Stine, a professor at San Francisco State University’s earth and climate sciences department, why Apple’s weather app sucked, he scoffed at me.
“Not knowing whether it’s going to rain in an hour? I would say that’s just being fussy about where the peas are on your plate,” Stine said. “It’s an incredible technological achievement to know that it’s going to rain at all this week. I grew up in a world where weather prediction was not accurate. We didn’t have enough data. But over my lifetime, the skill of weather prediction has increased pretty astoundingly.” Talking with Stine gave me a new perspective on my gripe with weather apps. When you consider how much better these predictions have become over time, these apps feel more like an achievement of technology, instead of a point of annoyance.
:format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/25025784/1708167204.jpg)
Stine explained that everything we think about forecasting comes from the National Weather Service, a branch of the National Oceanic and Atmospheric Administration. Every six hours they run a simulation which then gives them information for the next few weeks. Regional offices break down that information pragmatically, with attention to past data. Weather companies (e.g. Accuweather, Weather Underground, etc.) then go and make nudges and tweaks to that information to create predictions.
“Ultimately, whatever someone’s putting on an app — they don’t have access to different information than anyone else,” Stine said. “There is not different information available to different weather predictors. They’re all using the National Weather Service.”
The models, Stine says have improved greatly as time has gone on, getting better and better every day. That’s mainly due to more and more detailed data being fed into the equations over the years, to the point where there’s more uncertainty in current satellite observations than in the forecast models themselves.
The basic idea: everyone gets their weather data from the same place, and there shouldn’t be drastic variances between what weather companies and apps are saying. Also: stop complaining.
But Stine did have a small concession. He explained that my complaints aren’t about the grand scale of weather forecasting which, as he pointed out, can have major economic and governmental impacts. My grumble, he said, is more about the trend of what he calls “now-casting” — and that’s a very different animal.
“The traditional weather forecasting problem is a problem of understanding the fluid dynamics of the entire planet,” Stine said. “Whereas the problem is if it’s gonna rain in five minutes, that’s a very localized [concern]. That’s not something that, to my knowledge, the National Weather Service is very interested in. It’s kind of neat though, and maybe you can go put on a coat, or maybe you can go out back and stick the bicycle in the garage.”
Complaining about Apple being wrong about rain in Manhattan in seven minutes when, over Stine’s lifetime (he’s 49), there have been massive developments in weather prediction does feel a little like complaining about the way the peas have been arranged on my plate. How Stine thinks about weather forecasting and how I, pre-Stine, thought about weather prediction differed in scale and scope.
But those disparate perspectives find common ground when it comes to the importance of meteorologists.
As accurate as these models and forecasts are, meteorologists are key to understanding the weather around us, how it behaves, and the places we live. Apps will never, barring some kind of future, massive technological advancement, be as good at weather prediction as the meteorologists who understand how a particular combination of physics, mathematics, and geography work.
“It’s part of understanding the value of meteorologists. And this is not me, like, trying to defend my job,” said Homenuk, of NY Metro Weather. “Human input is needed to understand the complexities of weather.”
Homenuk told me, as of the time we spoke, that he didn’t expect any rain in the forecast for New York City for Halloween. I’d check the app, but I am gonna trust him on this one.
