
Sign up to save your podcasts
Or
Since I last mentioned weather apps when I did some spring cleaning on my iPhone, I've settled on a single solution that works best for me: Storm by Weather Underground. (It replaced their other, slightly less geeky "Wunderground" app.) But the fact of the matter is that, unlike a few years ago, I haven't tried many new weather apps recently. Patrick Dean summed this up on twitter, pointing out that just about any competently made app reaches the threshold to be featured, as weather is still a top-level category in the App Store.
I think there's one clear reason for this: homogenization of weather data. In the early days of the App Store, weather apps could differentiate themselves based on the source and accuracy of their data; attractive presentation came second. That all changed three years ago with the introduction of forecast.io. If you don't know the history, forecast.io is the open API by the creators of Dark Sky.
When Dark Sky came onto the scene in 2011, it was head and shoulders above other weather apps, with groundbreaking features like predictive radar maps and push notifications for impending precipitation. But even in version 1.0, it was clear to me that some things were amiss. It was less accurate on the iPad than on the iPhone because it made its weather predictions client-side, using the GPU. In fact, Dark Sky is a classic example of "machine learning applied to a problem we (and the machine) know nothing about". Its algorithm doesn't analyze weather data, it analyzes images, specifically radar map frames. This is what gives its future maps an unearthly, unrealistic quality; storms that have been developing or diminishing suddenly turn into zombie clouds drifting in straight lines across the map. In fact, if you click the play button on the map in the latest version of Dark Sky, it stops at the present time, as if it's ashamed of how bad those future maps look.
In serious contrast, Storm offers a 5-hour future radar map underpinned by meteorological data. Storms can appear out of nowhere along front lines, or die down before reaching your location. It's this one feature, plus its excellent daily and hourly graph views, that make it my top weather app. Almost everything else in the App Store is just the same forecast.io data — a computer's idea of weather — dressed up in new skins. The greatest innovation has been to make the interface super-snarky; an advance in entertainment, but not in utility. There may never be another great iOS weather app, because the hard part is gathering the data. Weather Underground operated stations for years, and is now (regrettably) under the weather.com umbrella. There is free weather data out there, but you get what you pay for: a slick interface for a couple of bucks.
5
22 ratings
Since I last mentioned weather apps when I did some spring cleaning on my iPhone, I've settled on a single solution that works best for me: Storm by Weather Underground. (It replaced their other, slightly less geeky "Wunderground" app.) But the fact of the matter is that, unlike a few years ago, I haven't tried many new weather apps recently. Patrick Dean summed this up on twitter, pointing out that just about any competently made app reaches the threshold to be featured, as weather is still a top-level category in the App Store.
I think there's one clear reason for this: homogenization of weather data. In the early days of the App Store, weather apps could differentiate themselves based on the source and accuracy of their data; attractive presentation came second. That all changed three years ago with the introduction of forecast.io. If you don't know the history, forecast.io is the open API by the creators of Dark Sky.
When Dark Sky came onto the scene in 2011, it was head and shoulders above other weather apps, with groundbreaking features like predictive radar maps and push notifications for impending precipitation. But even in version 1.0, it was clear to me that some things were amiss. It was less accurate on the iPad than on the iPhone because it made its weather predictions client-side, using the GPU. In fact, Dark Sky is a classic example of "machine learning applied to a problem we (and the machine) know nothing about". Its algorithm doesn't analyze weather data, it analyzes images, specifically radar map frames. This is what gives its future maps an unearthly, unrealistic quality; storms that have been developing or diminishing suddenly turn into zombie clouds drifting in straight lines across the map. In fact, if you click the play button on the map in the latest version of Dark Sky, it stops at the present time, as if it's ashamed of how bad those future maps look.
In serious contrast, Storm offers a 5-hour future radar map underpinned by meteorological data. Storms can appear out of nowhere along front lines, or die down before reaching your location. It's this one feature, plus its excellent daily and hourly graph views, that make it my top weather app. Almost everything else in the App Store is just the same forecast.io data — a computer's idea of weather — dressed up in new skins. The greatest innovation has been to make the interface super-snarky; an advance in entertainment, but not in utility. There may never be another great iOS weather app, because the hard part is gathering the data. Weather Underground operated stations for years, and is now (regrettably) under the weather.com umbrella. There is free weather data out there, but you get what you pay for: a slick interface for a couple of bucks.