I now use Weathergraph which does it differently but I would go back to Dark Sky (and pay for it) in a flash.
It shows the correct things and on a phone understands that showing the temperatures across the screen is useless as if I go out I want to know what the weather is like when I might make the journey back in 8+ hours time. I might not care what the weather is in 4 hours time as I will be inside.
The information design argument is 100% valid, but I also marvel that, having bought the company, Apple's weather app still isn't as precise or accurate. I don't know whether Apple's privacy focus prevents them making the same precise predictions, or if there is some other reason they don't, but it's sad that in 2025 we don't have the same level of performance as we did twelve years ago.
In all seriousness I heard some good things of dark sky. My current weather app is windy.com and I believe it's more built for surfers and such (??) - not sure what the best android weather app is.
So much of weather forecasting, at that time, was about trends and probabilities. DarkSky was about events, certainty, and action.
It was truly ahead of anything else and forced a new standard.
But iOS has this now. It's the same thing. They integrated it from Dark Sky.
This is why I don't understand the complaints that iOS precipitation accuracy is worse than Dark Sky's. The map works the same way. The chart works the same way. Complaints about UX I get. But not the complaints about a supposed fall in precipitation accuracy.
I get that it's a common trope that products always supposedly get worse once they're bought. But in this case, in terms of accuracy, I just don't think it's true. And remember, Apple would have zero reason to worsen the quality. The whole point of buying it was to improve iOS weather. Which it did.
Is it not? The rainfall-per-minute over the next hour on iOS seems about the same accuracy as Dark Sky had -- I used Dark Sky for years. It wasn't perfect but it worked well enough, same as iOS did after. You can even scrub the precipitation map predictions and they look the same to me.
I know the Dark Sky prediction accuracy was greatly dependent on where you lived -- this is something that was widely discussed back in the day. If you've seen a drop in accuracy, did you simply move?
And just anecdotally, Dark Sky was a delight to use. Apple Maps makes it a chore to extract the same utility from their app.
(And Apple notifications are a mess generally. I constantly have notifications for something yesterday only show up today. I'm not sure that has anything to do with Weather specifically, their whole notification priority system is borked.)
I think -- and I might be wrong, since this is from over a decade ago -- that when I first used Dark Sky, I ended up disabling notifications because it would constantly warn me of precipitation, but then when I checked the graph there was none because the model had since updated, and I wound up turning them off. So notification thresholds are probably something hard to get right, and what is appropriate for one geographic area might not be optimal for another.
I've sort of transitioned to using Ventusky and Windy to checkout the big picture stuff, then I make up my own mind about precipitation. I live in the PNW of the US and our terrain is so varied that forecasting services are kind of meh in general. They're decent for "it might rain for a while today" but anything hyperlocal tends to get bad because of the terrain in Oregon.
The Doppler radar that "live" precipitation comes from takes 4-6 min to complete a scan, and then obviously it takes a few minutes for that all to be ingested, update models, and push to devices.
The "live" weather from Apple (and when it was Dark Sky) has always been a prediction from about 10 min ago. And if it's raining where you are but dry six blocks to the north (as happens all the time), it's understandable why it gets it wrong.
Also I really like a tool called Forecast Advisor. https://www.forecastadvisor.com/ . It shows you the accuracy of various forecasting services for your area.
I use it whenever I travel. I don't stick with one forecast site because depending on the terrain/location their accuracy changes drastically.
Certain models are better for certain geographical features depending on the location. I tend to hangout around a lot of mountains and the difference in forecast models makes a huge difference.
Apple Weather is nothing like this.
That said, I'd still bet a dollar (that to be fair, I might lose) that Apple today is less accurate, and if they're just as accurate twelve years on, that's a fail as well.
Lived in the same general area (just outside a major metropolitan area) where I use DarkSky and now Apple Weather app.
DarkSky has better data vis and more reliable prediction. Apple Weather consistently over predicts snow fall amounts and many times I’ve had to use the Feedback to correct it on current conditions (e.g. raining when it says no rain or vice versa). I believe DarkSky had the same feedback feature but I never needed it this much.
Most of the time AW is fine but it’s less good to the point I’ve considered alternatives.
Scrolling through the Dark Sky screenshots, I can recognize many of the same things now incorporated with Apple’s. And Apple does offer location specific notifications of rain which I find to be pretty accurate, about as accurate as Dark Sky.
There’s largely a perception problem with Apple. People loved Dark Sky as an independent small app that worked well, before Apple took it and destroyed it. Now, even if Apple incorporated all of the same data and features, it still wouldn’t give that same spark of joy people had.
This is what I really liked about DarkSky. I didn’t have to read and understand the forecast, I could simply glance at it and intuitively have an understanding of the day’s weather. Apple lost this, and I think it is what gave DarkSky so much value.
Even without any text labels, you should be able to get a feel for what the weather is and how it will change:
- Hourly plots like Dark Sky, with everything (temperature, rain, AQI, weather conditions) in a single plot.
- The change in temperature visualized with both color and space. Space is obvious (higher -> hotter); color ranges from red for hottest to blue for coldest. All the visible plots share the same color-temperature mapping. So the gradient block to the left shows both the temperature range for that day as well as how it compares to other days.
- Finally, there is a weekly overview at the top.
Not a replacement at all for Android subscribers!
Huge bag of data for you to mess around with. I've started to use it to do my own weather forecasting instead of relying on forecasting services. Where I live has a radar gap(Oregon) and ridiculously varied terrain, so forecasts aren't great anyway.
Interesting, I think it's gotten worse over time. Even basics like what the temperature will be in a few days. It's consistently ~5+ degrees off on the low side.
The developer is very responsive, lots of UI customization (both app and widgets) is possible, and pricing is reasonable.
Still the best of all the weather apps though.
Both give you a huge amount of layers and datasets to mess around with. Windy recently changed their radar stuff, though, so it might be a bit confusing.
All the other android apps mentioned here have the same issue.
I might try to open an issue in their GH, or even a PR... A toggle for "denser graphs" and a setting for hourly resolution could do wonders.
It really sucks. Do they use the same data? I've noticed Apple Weather is substantially less accurate than Dark Sky. If Dark Sky told me it was going to rain in 10 minutes for 7 minutes, that's what was going to happen. If Apple Weather says it, well, maybe.
Apple says they use data from the weather channel, but this varies based on country. It used to say right in the app, but it seems like they removed that in favor of this link:
Then one day it just stopped working and it all went away. But it took apple ~18 months or so to kill it off, if I remember right.
But there was never any refund or whatnot.
I've used it daily since.
Windy.app is for wind based water activities. Windy.com is a data-heavy weather information site.
Nice things:
- Loads fast
- Nice vis of both today and week
- Can mouse over the visualizations to get precise readouts.There's a screen shot showing what it used to look like before gradients: https://github.com/Leftium/weather-sense
Some differences:
- Shows weather from yesterday for comparison
- All hourly plot trackers connected; not just the top one
- Includes AQI
- Sky color visualization (try scrubbing across dawn/dusk!)
- Non-precipitation colors approximate sky color (haziness)
- Temperature variation visualized both spatially and with colors
- Data source is Open Meteo
- Planned: 60 minutely forecast like https://openweathermap.org
- https://weather-sense.leftium.com/?n=nyc
n is short for "name" and uses the Open Meteo geocoding API[1].
* Adopt a colour scheme with similarity to the old BOM?
* Some way to store longer baseline movie animations in local state so people can avoid cost in you but run the weather radar for longer?
* Tide info? Hyper specific to people who do water things. Willyweather does this really well.
I use Willyweather and Windy. I used to use a weather app written by some mob called "shifty jelly" and their git logs were .. hysterical. Drunk fairy penguins seemed to cause most of the bugs.
A website dedicated to data visualization and it's totally broken on Desktop Firefox. If they had just created a straightforward article, it would be perfectly legible, but all the flashy-flash just makes it unintelligible.
I have learned to ignore its predictions. It will say that it's sunny outside, and I'll look out the window, and we're having a hailstorm.
I've learned that I just want to look at the radar. There's a big difference between "it's going to drizzle all day" and "spotty storms within 25 miles of you"
- The data is from https://open-meteo.com
- It would be trivial to connect the historical weather API (back to 1940): https://openmeteo.substack.com/p/processing-90-tb-historical...
Polaroid
Pebble
Palm
Oldsmobile
Tower Records
Borders
Pan Am
Whenever I travel I find it pretty helpful. Certain services are just garbage in some areas.
For example, Foreca is like 84% accurate for my home location, but it's only 60% accurate for one of the cabins I frequent.
The table of user "context and situation" is a great document. You can easily envision authoring this table and scrolling to the right of your initial columns (A,B) to see further into the design process,
A) "When I hear about a storm, I want to prepare my loved ones, my property, etc.
B) Storm forecast ... : - Where is the storm right now and is it heading my direction?
[...]
N) _Show the storm front using _directional arrows_ ... (compact and replaces need for animation)_
The last section concludes in praise of the design and includes this: _"rigorously iterated on data visualization design". I wish we would have seen evidence of this, principally in the form of older screen shots of the design.
I think design iteration is the difference between mere good design and good products, and legendary product design.
Personally, I'd love to see a write up of my favorite whipping post, Transit App. Oh boy did that app go down hill, and with such great potential.
Weather APIs are pretty open. What's stopping you?
Check out their compare feature. Brilliant.
Eg: Seattle vs London https://weatherspark.com/compare/y/913~45062/Comparison-of-t...
(Not affiliated. Just an admirer.)
In all other places in the app, the low is to the left of high.
The idea that Apple is full of good designers should be forgotten. They're as mid and sloppy as any other large tech company.
During a cold snap one cold night will show up as the low for two consecutive days instead of a single “overnight”
This topic raises an issue I’ve had in mind for a while, companies are not realizing their true value when they sell out to some exit, as is evident by the fact that the companies Andy what they created end up being taken out behind the shed. If a competitor is willing to pay a certain amount without extreme pain to the point of convulsion or you don’t get air tight contract that prevents killing off the service/product without it remitting back to the founders or being made open source, you are being low-balled.
Taking the Wunderground example, those folks would have ended up owning the weather channel and probably buying or merging with Dark Sky and being the data provider to Apple instead of the Weather Channel characters (in case you don’t know about that entity) owning and killing off their baby.
they've been lobbying for like a decade to get NOAA defunded. They're basically the Intuit/turbotax of the meteorology world.
One of the things that I've seen with them that I haven't seen with others is the cloud cover by layer.
https://www.yr.no/en/details/graph/2-6301678/United%20States...
https://www.yr.no/en/details/table/2-6301678/United%20States...
For doing photography (sunsets) there's a significant difference between 50% high clouds and 50% low clouds.
A year and a half or something later.. I recently started a project of my own trying to bring all "weather dependent" photo opportunities together in one place, if you wouldn't mind I would be happy to experiment with bringing Sunsethue data to https://photoweather.app - your prediction model is certainly a lot more sophisticated than mine and it would be very cool to offer that
It was a rare example at the time when it was _the_ webapp better than any existing 'native' apps.
Look at image. Scroll down to find the next image button. Scroll back up to look at image. On desktop
So it would be like "60% chance of rain after 2pm, total amount less than 1/10th of an inch"
It tells you all you need to know at a glance :)
e.g. see https://www.meteoswiss.admin.ch/images/1904/website/weather/...
(from https://www.meteoswiss.admin.ch/weather/weather-and-climate-...)
I've noticed there is a correlation, but having both is useful:
- Often there is a percentage chance, but the mm/hr is 0. At these times, it could rain but will probably be very light.
- Less common, but sometimes there is 0% chance, but a non-zero mm/hr.
I think that would make a great single purpose mobile app - automatically knows where the sources of information are and shows you the rain - where it was, and where it is going.
A eulogy for Dark Sky, a data visualization masterpiece (2023) - https://news.ycombinator.com/item?id=41109799 - July 2024 (196 comments)
A eulogy for Dark Sky, a data visualization masterpiece - https://news.ycombinator.com/item?id=35263115 - March 2023 (251 comments)
(and yes, the visualizations were beautiful, but the real key was being able to see exactly where one was with respect to the radar picture and to be able to use already existing forward predictions of the radar picture in conjunction with precise gps to generate timeseries/events.)
I've settled on using the built-in Android weather app, but it pales in comparison to Dark Sky, in every respect.
I don’t know if Carrot can use Apple Weather as a source on Android. Also it seems like the Android version is not in active development.
Toggle only the stats you're interested in! The toggle is persisted to localStorage.
I plan to add more stats, like wind speed and direction, but they will all be toggle-able.
It's shown in the middle screenshot at https://weathergraph.app (on desktop, mobile users can check https://impresskit.net/image-download/9161183f-e118-4c75-8f8... )
So true.
Open Meteo supports 28 different WMO weather condition codes[1]. Most weather apps only support half as many. (Just "rain" instead of light/moderate/heavy rain.)
Showing all 28 is less helpful because of the noise. More useful just to show it might rain for a period of several hours vs oscillating between light rain and heavy rain. The light vs heavy precision wasn't worth it when there was high uncertainty whether it would even rain at all.
So https://weather-sense.leftium.com consolidates hours with similar weather conditions into a single segment by default. You can click on the weather icons at the left of the plots to toggle the original unconsolidated view.
Wetter: http://plot.micw.org/apps/wetter/index.php
weatherstrip : https://www.weatherstrip.app
Eg;
https://forecast.weather.gov/MapClick.php?lat=42.3773&lon=-7...
Specifically: each day has a range (low and high) but it's not clear whether the low is for the morning or evening, and they could be vastly different. You could have 10-15 one day then 0-10 the next day, and think "Ok, I'll go out tonight and bring a jacket but no hat since the lowest it'll get today is 10 and whoops, actually it's freezing by the time dinner's over.
There are so many ways apps could do this better. Like showing a vertical line graph rather than discrete bars, with the lows inbetween days. Or if you want to keep the bars, make them angled, so the low is closer to the morning/night it's associated with. Or even show 3 temperatures, not just two! (one being the low for the previous or next day or whatever)
mocmoc•17h ago
FredFS456•17h ago
Hnrobert42•17h ago
hdx19•16h ago
Hnrobert42•12h ago
fsh•17h ago