Join 3,377 readers in helping fund MetaFilter (Hide)


Why does WeatherBug's data suck, and is there anything better?
December 15, 2013 9:14 AM   Subscribe

Last night I was checking WeatherBug, the iphone app I use for weather, and it said the expected low for the night would be 25. It was also telling me that the current temperature was 18. In the forecast for today, it predicted a high temperature of 42. Great! Snow melt! So I checked the hour-by-hour forecast to see when it would hit that high, and it didn't show anything higher than 35. This sort of thing happens a lot, and it offends the nerd in me. So: can other apps do better, or do they all just feed off the same NOAA data?

Another issue: raw data is often not enough. I tune to local radio weathermen to hear not just the most likely outcome, but the confidence level, and the best/worst case scenarios. That info requires a human touch, and none of the weather apps, understandably, provide that. Or do they? Are there weather apps which do include some subjective human judgement and detail in the loop?
posted by Quisp Lover to Computers & Internet (23 answers total) 8 users marked this as a favorite
 
I've found Forecast.io awesome, though they don't have an app per se (I just keep a hyperlink to their page on my phone's home screen). From the same devs is Dark Sky, so that might be worth a try (I'm an android user, so don't have direct experience with that one).
posted by Emperor SnooKloze at 9:20 AM on December 15, 2013


I've used a few weather apps, and I also tend to use Forecast.io. They do their own calculations on available weather data, which been pretty accurate in my experience. I recently also read some good things about Weather Line, which I believe uses the same data.
posted by Magnakai at 9:23 AM on December 15, 2013


I just want to confirm this is a problem I've noticed with many weather services. It's not just a WeatherBug issue.
posted by 2bucksplus at 9:27 AM on December 15, 2013


I've tried a few weather apps and honestly the Weather Channel one is the most accurate (for my area at least) though I'm pretty sure it also uses National Weather Service data.
posted by rabbitrabbit at 9:28 AM on December 15, 2013


I find Weather Underground to be much better where I am than other services. They use data from people's local mini weather stations, as well as the nationally available info, and it helps a lot.
posted by limeonaire at 9:38 AM on December 15, 2013 [7 favorites]


FWIW, it looks like there is an iOs app for forecast.io (not sure how official it is, but it does use their data).
posted by Quisp Lover at 9:39 AM on December 15, 2013


Sometimes if you're in a weird location relative to your local reporting stations, you can get bad forecasting. I live in a weird valley between three non-valley weather stations and I get that problem sometimes. I really like WeatherSpark because it gives me a set of forecasts that I can triangulate among and a lot of good time-based forecasts as well as recent weather history. I've found it matches what I see outside my window decently. Right now there is only an Android app for them but their mobile interface is pretty nice.
posted by jessamyn at 9:44 AM on December 15, 2013


Weather Underground has new versions out for both iphone and android -I haven't used either of the new ones but have generally found the site to be pretty accurate - and full disclosure -am friends with the people developing both apps.
posted by leslies at 9:48 AM on December 15, 2013


Attention, thread:

FWIW I'm not looking for more accurate forecasts. I'm pointing to an issue of internal consistency in Weatherbug and asking how and whether that can be avoided in other sources.

I suspect people are reading the headline but not the question.
posted by Quisp Lover at 9:50 AM on December 15, 2013


As Emperor SnooKloze stated DarkSky is the official app made by the forecast.io developers, the app you linked is unofficial.
posted by Harpocrates at 9:51 AM on December 15, 2013


I user WeatherUnderground. Pretty accurate. Seems to predict rain a little too often, but I'm a builder, so a little extra caution is welcome.
posted by MeanwhileBackAtTheRanch at 9:53 AM on December 15, 2013


I really like the Yahoo Weather app. It's visually very nice and I haven't seen any inconsistencies in the data.
posted by joan_holloway at 10:13 AM on December 15, 2013


I'm not sure I see your issue. The EXPECTED low is not really the lowest temperature actually achieved and was probably computed at 6:00AM for the entire day. Similarly the expected high is a prediction for a peak and therefore is a single number - hourly averages need not actually match the high (especially if it only peaks for a short period of time).

I understand the intent of your question, and your frustration but you may find that all sites have this problem, but not because they have bad data because they are reporting something which doesn't exactly match what you would like to see (i.e. lowest achieved temperature, finer grained forecasts....)
posted by NoDef at 10:54 AM on December 15, 2013 [3 favorites]


Internal consistency is just not what weather forecasts have; if you believe the poor meteorologists who are called to defend their profession in the media after an especially gross error, the only consistency there is is that many weather formations do unpredictable things. It's pretty much that the "forecasts" are reliable only shortly after the fact, which makes that the whole enterprise, seen from the consumers' end, is and remains a bit of a joke.

That said, different forecasters use different combinations of data, and so it may seem that some are more accurate on average than others. This may differ depending on where you are, and also it may be coincidental after all.

And THAT said: here's the one I use because it's less random than the official Royal Swedish pain in the forecast: https://itunes.apple.com/gb/app/yr.no/id490989206
posted by Namlit at 10:54 AM on December 15, 2013


Forecast.io lists their data sources, which seems to aggregate from a variety of sources.

Weatherbug doesn't seem to be that straightforward with their sources.

wunderground offers two forecasts for the US, their own proprietary forecast and one from the National Weather Service.

Both weatherbug and wunderground, and probably others seem to pull current temperature data from personal weather stations, which you can buy and hook up to feed back your current weather conditions to them. Maybe weatherbug has inaccurate weather stations, which is skewing their forecast reports?
posted by hooray at 11:14 AM on December 15, 2013


anecdotally, I have weather bug on my phone and use weather underground on my desktop. the Bug is often very buggy, and I know what you mean: yesterday it snowed all day here, yet the "now" cast gave an 80% chance of snow. really??
posted by Dashy at 12:51 PM on December 15, 2013


Dashy, yes, that's what I'm talking about. So do you see that sort of thing with weather underground?
posted by Quisp Lover at 1:18 PM on December 15, 2013


Weather Underground's forecast high and low often differs from the highest and lowest numbers in the "hourly" forecast, but I think that's because their "hourly" forecast is really "every 3 hours." For where I am, I've found Wunderground's own forecasts to be more reliable than the NWS.
posted by brianogilvie at 2:34 PM on December 15, 2013


Another vote for Weather Underground. I've been using it for years, and comparing it to other sites/apps that have come along. It continues to be the most reliable for me.
posted by trip and a half at 3:24 PM on December 15, 2013


Emperor SnooKloze: "I've found Forecast.io awesome, though they don't have an app per se (I just keep a hyperlink to their page on my phone's home screen). From the same devs is Dark Sky, so that might be worth a try (I'm an android user, so don't have direct experience with that one)."

For those using Android, there's a forecast.io app called Arcus - works well for me.
posted by namewithoutwords at 5:04 PM on December 15, 2013


weather forecasting services use different forecasting models and it can affect their accuracy. Additionally some are only using a few data points (airport weather stations) and some are using many local private consumer weather stations. Weather Underground tends to be one of the more accurate services because they do a mix of using NWS data and "crowd sourcing" their data from local users with personal weather stations. In addition, the WU app now allows you to "verify" that the weather they're reporting is correct, and if not, you report what's off and why.

It is by far the most accurate service I've seen. I have the iOS basic weather app, Weather Underground app, Forecast.io iOS app, Yahoo! Weather app, The Weather Channel and WeatherBug on my phone. I use WU probably ten times as often as any of the others; forecast.io is my second choice, mainly because they tend to be more accurate on precipitation (but not temperatures, probably because they don't crowdsource like WU does).

I would concur with you that WeatherBug kind of sucks. Their iPad app is gorgeous but they rarely get the temp forecasts right for our region, much less precip. Although I'm kind of used to that because when you live this close to 14,000' peaks, a slight change in wind direction can raise or drop the temps 5-10ยบ in a few minutes.

For the most in depth weather for my region, I actually use a specialized ski weather forecasting website called OpenSnow.com. They are strictly focused on ski weather but since Joel Gratz (the guy who runs the site and does the Colorado ski forecasts) happens to be a Boulder resident, it means we here get very good accurate in depth detail about not just the short and midrange forecasts, but also the dynamics that create weather patterns here in the Front Range. It really is very, very difficult to forecast here; fluid dynamics are a bitch in steep terrain, and Joel is a HUGE weather nerd who is just as excited about when he gets it wrong (which is actually pretty rare) as right, plus he's a great teacher and explainer of why weather works the way it does. If you have any interest in the dynamics of snowy mountain weather / ski forecasting, go check out his blog - it is fascinating.

One of the things that Joel has said over and over and over again is that the way he does weather forecasting is mainly based on modeling data from the big machines in national/international services, and these various models are varying levels of accurate, depending on circumstances. Weather is tremendously complex. He tends to prefer the European forecasting models, with a side of Canada's service, because they're by far the most accurate, at least for our region. Forecast.io, I believe, bases their forecasting on the EU models as well. They're newer modeling tech with more powerful computing behind them. The American GFS models, in comparison (which the NWS uses) are a bit older and all signs point to them tending to be somewhat less accurate. Oftentimes the ski forecast guys will run averages of 3 or 4 separate modeling services to see if they agree and how well they align.

If your weather app / service uses the GFS models (NOAA, NWS, Yahoo, WeatherBug, The Weather Channel, most other US services) then they'll only be as accurate as those models. If they're only using a few scattered data reporting stations (official stations / airports / etc.) then they're not going to be quite as locally accurate or granular as a service that dynamically updates / "crowd sources" their data like Weather Underground.

So that's a lot of words to explain why the weather is so tough to accurately forecast.
posted by lonefrontranger at 7:10 PM on December 15, 2013


Incidentally, Forecast.io has the same kind of issues. For instance, if I look at it right now, it says today's temperatures will run from 16 to 27 degrees. But if I expand the hourly view, at 2 PM, it predicts 28 degrees. Same thing tomorrow.

I think the core issue is that apps consider the predicted high and low to be immutable, and they aren't updated based on data from today. Useful if you want to test the prediction after the fact, not useful when they are clearly wrong at this very moment.
posted by smackfu at 8:25 AM on December 16, 2013


Are there weather apps which do include some subjective human judgement and detail in the loop?

For information beyond just the numbers, I find the area forecast discussion links at the NWS to be instructive.
posted by doctord at 2:59 PM on December 17, 2013


« Older I have a good life - good comm...   |  A month or two ago I read abou... Newer »

You are not logged in, either login or create an account to post comments