Last night I was driving to NYC. On my way to the George Washington Bridge there was an LED sign above the highway saying that there would be snow on Monday. When I got to my apartment around 9:30 pm I opened the Apple Weather app and it had a 0% chance of precipitation for Monday. I checked again this morning around 8 am and it had changed to a brief period of snow around noon, but stopping within the hour.
In reality it started snowing around 10:00 am and at 1 pm it is still snowing. From what I remember from Dark Sky, their precipitation forecasts were pretty good. I know Apple bought Dark Sky, are they no longer using the same data source?
Yesterday, when I woke up, it said the “heavy snow” would be ending in 30 minutes. When I looked outside we were receiving light rain. There was no snow in the county. Dark Sky used local weather radar to track precipitation in real time. That was the DK’s best feature and Apple Weather doesn’t use it.
I rate AW as slightly better than the Old Farmer’s Almanac.
When I posted this just after one, Weather said snow stopping in 23 minutes, which is mysteriously precise. Of course 50 minutes later it was still snowing.
Take a look at windy.com (NOT windy.app). It has 40+ map layers that track “everything” including big fires and radiation. The app and website are free to use and the premium version gets more up to date info, etc. I use the weather radar to predict when weather will arrive.
Weather forecasting is hard. Dark Sky did the simplest version, which was looking at where it was raining now and predicting where that rain might move in the next half hour. It was often accurate, but not always.
In the last couple of days weather radar has often looked like it was raining here when a glance out of the window showed it was dry. Rain was falling, I could see lots of it below the clouds, but evaporating into dry air before it ever reached the ground. Once it made the air humid enough it reached the ground and carried on raining enough to flood low places hereabouts.
There is no substitute for a combination of extremely fine scale measurements and mathematical forecasts, interpreted by people with local knowledge and experience. That costs, a lot. It used to be seen as a government obligation to citizens to provide a single, authoritative set of forecasts but the last couple of decades have commercialised the whole field so you get what you pay for. Air Traffic Control forecasts (METARS and TAFS still paid for by government) are remarkably accurate in the UK at least but we’re at least 50km from the nearest airport, so they won’t work for us.
In NYC it said we would have < 1in for the last few days and that is what we got. This was not a big miss. A few days ago it had 1-3. In the normal range for snow forecasts.
I appreciate Dark Sky may have worked for some, but in my view (really, my location), ALL weather apps are to prediction what a wet finger in the air is to observation — a general idea of things. Because…
People often tell me “look out the window” but for that to work, I need to move house, as although my office is only 12km away, the weather is often very different there due to local terrain. So I have to use weather apps and guess at how accurate they’re going to be.
Your value judgement is way different to mine. Less than 7c per day for something you will use most days seems very cheap to me.
I suspect we aren’t that far apart. I’ve been a subscriber for a couple of years, but my statement was a warning for those that think $0.10 a day for 1Password is expensive.
Windy is definitely an app for weather geeks. Multiple forecast sources are available, and can be compared. The web app is the best interface. The phone and tablet interfaces are OK, but not as easy to navigate, IMO.
I live in NJ and commute into NYC often. For yesterday’s storm, I checked Apple Weather on Sunday to determine if I should commute into NYC on Monday and it initially said 1" in my area and then later in the day changed to 2" of snow and that is essentially what we got. I ended up going into NYC yesterday - with no issues on my commute - due in part to the info from Apple Weather. And when I was in NYC yesterday, Apple Weather was fairly accurate to what was happening outside. As noted, predicting the weather is damn hard and while I miss Dark Sky, I’m pretty happy with how Apple has evolved its Weather app to use the key elements of DS.
This is not a dig at you, but an opening for me to make a dig at (almost) all the reviews of weather apps. They focus on the USA. Sometimes Europe. I live in an island country in the Southern Hemisphere, in a part that has unique weather-affecting geographical features (usually to our advantage). Most weather app reviews, at least when it comes to the data aspect, aren’t worth the bandwidth used to consume them.
Very hard because they make gross generalisations by location. Per my comment above that I live 12 km from my office. The local weather observations are taken quite close to my office so bear little resemblance to what I get at home, but the official state forecaster has an app that tells me the weather for my specific locality. Well… they say that. I have checked on many occasions. There must be several dozen “locations” it will forecast for that are all always identical.
To give a real example about how hard it is to describe current weather. I live in a town on a peninsula that sticks out into the English Channel (and really the Atlantic). The other side of the peninsula is an escarpment that’s about 500 feet above sea level.
There’s often a difference in windspeed between my house and the lighthouse at the end of the peninsula (6 miles away) and in a Westerly it can be Force 10 or 11 at the lighthouse and Force 4 here. That’s the difference between mildly windy and having your roof in pieces. In better weather, sea breezes mean that the wind here can be 180 degrees from the prevailing wind. We’ve a risk of snow here today. It’s very common for the town not to see any, but even the main roads over the escarpment ( maybe 4 miles away) to be completely blocked and by feet of the stuff and dangerous. We might get two days of frost in a winter month, but folks on the escarpment might get 10.
Almost all the weather forecasts and apps give this town and the one ten miles away to the north or 20 miles each way along the coast. A few have the lighthouse as it’s an important navigation point for sailors. An accurate forecast for this area could be correct to say it will and won’t rain today, it will and won’t be extremely windy or that it might have snow and definitely won’t have any.
You have to interpret what the forecast models show. I know what the sequence of weather is likely to be if a storm is approaching from the Atlantic, or if High Pressure is building and that might allow me to “model” the forecast: what won’t be accurate, and could be very misleading, would be relying on things like a 60% chance of rain in the area that has been produced by the forecasting model.
I’m interested to see LLM-type AI applied to this. Pattern matching (when the model says this the actual weather is that) is what they are good at, but I wonder if Apple Intelligence, or a competitor, is ever going to build a model that is local enough to be useful.
