Weather apps: a tricky topic; most people wouldn’t understand my obsession with weather apps and widgets. But I had to move the section about weather apps out of my post regarding the Android apps I’m using and recommending, simply because it kept growing, and growing…

The short version: they all suck. I’ve written more than once about this, and I deleted the respective blog posts. The longer version: there are more aspects to be considered.

Since I change my main weather app every once in a while–because, as I said, they all suck!–and because there is no such thing as “the best weather app” and one should determine what’s the most decent weather app for their location, what follows is a lengthy discussion on the main weather data sources and the available apps.

[01]—People who live in the US might be happy with a lot of weather apps designed with the US in mind, especially as some such apps have a tendency to prefer weather readings and forecast from airports (METAR). People living in Continental Europe are bound to be disappointed, unless they live very close to a very large city or to an airport. (Side note: using airport readings needs caution with regards to air pressure, as the METAR data only includes the QNH pressure, i.e. the pressure adjusted for the sea level, not the actually measured one.)

[02]—Most national weather services have too few weather weather stations, and the apps that use weather sources based only on official weather stations plus airports weather stations (e.g. eWeather HDGismeteoForecaWeather, MSN Weather–which uses FORECA) are not “fine-grained” enough (to make things worse, when FORECA doesn’t receive current data from the selected weather station, it will continue to show temperatures that are hours old!).

[03]Apps that prefer to use the closest airport even for large European cities that have better weather stations (e.g. WeatherBug, 1Weather) are simply useless! With Weather & Clock Widget Android one can opt between FORECA and MyWeather2.comMyWeather2.com will only use METAR readings from airports, e.g. for everything around Stuttgart the choices for the current conditions are EDDS/STR (Stuttgart-Echterdingen Airport, civilian), ETHL (Laupheim Air Base, military), and EDSB/FKB (Karlsruhe/Baden-Baden Airport, civilian)!

[04]—The finest granularity stays with apps that believe in the “hyperlocal” approach of using the tens of thousands of independent PWS (Personal Weather Stations), such as Weather Underground and any app that uses it. The problem is that such PWS are of extremely unequal quality: the other day I could find on a radius of 10 km dozens of PWS showing any temperature between -14°C and +12°C! In the official app one can select a different PWS than the one chosen by default for a given location, but in other apps that use WUnderground such thing isn’t possible (it used to be possible in Morecast, now not anymore; it’s also impossible in Weather Timeline; and MacroPinch’s Weather is so devoid of features that one couldn’t ask for such a thing); but even if I were to choose, I wouldn’t trust any of the PWS I know of! Instead of 180,000 contradictory and unreliable PWS, a smaller number of trustworthy PWS would be preferable. Surprisingly enough, both Morecast and Weather Timeline use rather decent PWS, although apparently not necessarily the same (it might have to do with the way my location is interpreted–Leonberg or Eltingen).

At some point I’ve determined that, for 71229 Leonberg, Morecast uses ISTELTIN3 (SW of Eltingen), a simple Netatmo Weather Station that is often broken (another PWS is chosen when this happens), while Weather Timeline makes a better choice by using ILEONBER11 (August-Lämmle-Weg, Eltingen), or sometimes ILEONBER18 (Adalbert-Stifter-Straße, Eltingen)–just like the official Weather Underground app. (Eltingen is part of Leonberg.)

Using Weather Underground can be at times perfectly satisfactory. Here’s its 3-day widget, right under Yahoo’s hourly one:

But what to do when the temperature has raised to about 7° (Yahoo still shows 5°), yet the app “believes” there are 13°C?

[05]—If you’re still unconvinced that Weather Underground’s PWS are unreliable, take a look at this map:

Any temperature between 23°C and 34 °C, really? Or, in a small town (right), 24 °C and 31 °C, or 23 °C and 29 °C only a couple of streets away? Such PWS are installed just out of the window, sometimes right under the mansard roof, and they receive direct heat either from the sun or from heated roof tiles! During the winter such PWS would also show abnormally high temperatures, under the influences of the same dwellings. There is no way their readings would be trustworthy!

[06]—Since I mentioned Yahoo! Weather… Funny thing, it happened to be the weather data source for the built-in iPhone app prior to iOS 8. Under the hood, it’s said to be using data from The Weather Channel, and weather.yahoo.com indeed is showing videos from The Weather Channel, but the readings and forecasts are slightly different, and less reliable than those displayed by weather.com. It tends to be optimistic about the chances of rain (e.g. showing 35% when others say 60%, or 0% when others say 30%), unless it totally fails to predict a rain. Either way, the widget icons will show rain starting from a chance of 40%. It’s definitely not suitable for some locations such as Brașov (Romania), for which it seems to toggle between a correct and a wrong data station. When a widget fails to update, the small update symbol works even under Marshmallow or Nougat. The developer of Chronus complains about the lack of reliability of Yahoo’s weather API, but there are some alternate query choices too. I typically prefer the hourly widget. Note that on dog days, Yahoo prefers to downplay the issue, so it will forecast e.g. 28 °C instead of 33 °C, so don’t rely on it during a heatwave.

[07]—The widely acclaimed Dark Sky, formerly known as Forecast.io and used as a data source by other apps too (Amber Weather and their ton of crappy widgets masquerading as separate apps, Arcus Weather, Forecaster, and as only one of the sources by Weather Timeline, Today WeatherChronus, and the newer apps Chill Weather + WidgetsHello Weather, and Mosum), claims to be “hyperlocal” too, and it might be accurate enough for US and UK, but for the rest of the world it’s mostly a hit-or-miss situation!

There is a simple explanation to that. Their data sources are almost all forecast models offered by NOAA, with the exception of the UK and Ireland who can benefit of UK Met Office’s forecasts and warnings; the rest of the planet relies solely on NOAA’s Global Forecast System, and NCEP’s Canadian Meteorological Center model, but NCEP belongs to NOAA, so even the Canadian model comes from a US source! It’s no wonder then that the forecasts are coarse approximations outside North America and UK: sometimes they’re close to reality, sometimes they’re ridiculous, and they lack the “hyperlocal” aspect, being only computations made on maps, with no input from actual weather stations.

And yet, the official app cannot be downloaded from Google Play globally: “Dark Sky is currently available in the US, UK, and Puerto Rico.” Alternative sources exist though: both legal (original APK) and illegal (patched APK to look as subscribed).

The forecasts and map-based computations have however greatly improved mid-2017 even for outside US/UK: the global worst-case resolution has improved from 40 km to 20 km, and NOAA’s High-Resolution Rapid Refresh model is now used. The long-term forecasts still seem kinky if you’re outside US/UK.

Forecaster (so far the best-looking app that uses Dark Sky’s data) has a nice color code, greatly informative at a glance:

Still, there is no widget that shows the forecast for several days at once; and the official app has ugly widgets too (it was designed with iOS in mind). And for the “current readings” (being them forecasts for outside US/UK), Forecaster almost always lags behind the official Dark Sky app!

And there’s one more simplifications in Forecaster as opposed to Dark Sky, see the way the chances of rain for Thursday and Friday are depicted in each app:

Chill Weather + Widgets is a newer app, with graphics that could support some improvement:

The summary for a day only shows the maximum temperature, there’s no minimum; and there’s no percentage for the chance of rain. The app has two transparent widgets, 4×1 and 2×1:

Hello Weather is shown with mucho detail at [51].

Mosum is ad-free and open source, but it has no widget, and I find it confusing how the details for a day are as detailed and precise as those for an hour (really, a wind of 2.61 km/h for the day, as well as a fixed pressure?):

[08]—Back to general considerations, there are apps that rely on a smaller set of PWS, typically part of some large private networks, see e.g. MeteoGroup with its official app WeatherPro and official site wetter24.de–there is a discussion on MeteoGroup-based apps below, where I discuss the German apps; consult the list of weather stations. In theory, such apps have the opportunity of being good, but it’s not always the case.

[09]—There are apps or even major data sources for which I can’t tell the exact approach, nor could I identify the exact weather station used for my location. The classical example is The Weather Channel (weather.com, The Weather Company, LLC), which didn’t alter its choice of weather stations not even the acquisition of Weather Underground, Inc. The results are decent enough (the widgets aren’t anywhere near decency, except that they have now added a a new, resizable widget), and its forte is that when there’s snow accumulation it shows detailed forecasts and maps of the snowfall, but they have this idiocy of showing for any given day a pair of temperatures that are not the minimum and maximum temperatures for that day, but the maximum for that day, and the minimum for the night that ends in the next day’s morning! How stupid is that? Most European people would believe the shown minimum to belong to the early hours of the respective day!

Here, the minimum shown for Thursday belongs to Friday’s early hours, and the only reason Thursday is still shown at almost 1 AM on Friday is because the minimum for Friday… belongs to the previous day:

My other major complaint about TWC: why the Windows tile can show a 5-day forecast, but there’s no Android widget to show anything but the current conditions? This is preposterous!

Yes, the Windows 10 tile is blurry, but at least the information is there.

UPDATE: It looks like the update of Sept. 8, 2017 introduced a widget that shows the weather for 5 days, but only on tablets. This was a disruptive update, several users complaining that the app now opens in portrait instead of landscape–it’s like the tablet were a smartphone! The same version 8.1.1 doesn’t add the said widget on my smartphone though. I know the tablet version of the app, it’s really a different app, but now they’ve removed the tabs from its UI, which they plan to further refine in Oct.-Nov. 2017. Unfortunately, I couldn’t find anywhere the tablet version of the APK.

I want to clarify something: no matter how hard I’ll criticize The Weather Channel below, it’s generally a fairly accurate source, with very good forecasts!

[10]Google News & Weather uses weather.com too, and the widget is quite fine, but I cannot properly read the forecasts, because for any single day I have to look at the previous one’s minimum. One can check the hourly forecasts in the app though, but the display of the hourly temperatures is sometimes buggy. The forecast is unfortunately limited at today + the next 3 days, and I’d rather prefer 5 or more days.

A bug shared with the official app (The Weather Channel) is that sometimes the current conditions and the forecast are slightly different between the automatically detected location and the same location, but manually entered. It’s as it would choose a different PWS, or as if it would display the weather for two different towns that are nearby (in my case, for the north of Leonberg it could show the weather for Gerlingen, and for Leonberg-Eltingen, the weather for Renningen; in both cases, the named towns are 3…8 km away from Leonberg, but in opposite directions.)

And there isn’t much to customize about the boring widget:

[11]—The youngest weather app–released in 2017–Weather Wiz, is surprisingly using data from The Weather Channel (“IBM weather data”), but at version 1.0.0 it’s so terribly buggy that most of the time it shows random data for my location (I wrote two detailed e-mails to the developers); refreshing the data in the app doesn’t mean it will display updated data (I’m not even sure it displays the data for the correct location!); also, the widgets need some fixes too.

My take is that the app does not query “IBM’s servers” (weather.com’s API), but rather their own cloud server, which seems to have two problems: it doesn’t update too frequently the readings for a given location; and it mixes up the IDs of the queries coming from the client apps, so that one’s app might read the data belonging to another location. I’m almost certain the random data it displays sometimes is due to the botching of different user sessions. Unfortunately, the stupid Chinese Canadian that answered my e-mails wasn’t even able to understand what the bug meant, and they were fucking suppose to know better their system’s architecture.

[12]—An “outsider” is Apalon’s Weather Live, as it claims to be using its own weather network (which is probably a lie; they only make apps!), and only the maps from FORECA. They’re a bit fishy, declaring an address in Dublin, Ireland on Google Play and Minsk, Belarus on their own website. Thankfully, their weather data is quite bogus–and I hate their widgets. They also have an app called NOAA Weather Radar & Alerts (translated in other languages without NOAA), which is even worse for locations outside US: in my case, during summer it constantly showed temperatures 2…4 °C higher than normal, as if they were using PWS situated on a brick or concrete wall!

[13]—ILMETEO’s app is pure crap, as most non-major locations are not recognized, and when they are, the data seems phony. Instead of Leonberg, it locates me in Gerlingen (because it doesn’t know what Leonberg is; it only knows the other Leonberg from Bavaria), and the shown temperature is almost always off by several degrees (say, 4°C different than any other app).

Another Italian app, 3B Meteo, is also crappy for locations outside Italy: some well-known locations must be searched in Italian, e.g. Londra, because London will find other locations; however, the app knows both Stoccarda and Stuttgart (the one from Germany, not from Arkansas or Kansas).

[14]—The French La Chaîne Météo / Weather Crave is even worse, with zero usability and useless data. See [45] for details.

[15]—Although I mention the peculiar Morecast more than once throughout this rant, I never quite liked it, although it’s packed with features (and ads). They originally allowed you to select the PWS if you wanted to–and they were the same PWS known to Weather Underground–but since some time they don’t reveal their sources for current readings; they only mention some large or national ones that could as well be used for the weather maps:

The app itself has an original design which however is not my cuppa tea. It also has several bugs, e.g. after granting it location rights, it still asks you to enter a location by name, otherwise it simply wouldn’t start. But searching for Leonberg gives four different “Leonberg, Germany,” with no indication on which is in BW and which is in BY; not to mention that there aren’t four such places anyway; so I normally give the app a different initial location (say, Renningen, for it’s unique). Strange enough, once the apps’s main screen is shown, one could select the properly detected location. The UI is a bit tiresome to me, but they stick to their graphic design. But maybe you’d like more their website.

Some of their widgets are not bad though, if it weren’t for the stupid icon in the top-right corner, and the slim typeface:

As a matter of fact, the company behind it uses two different Google Play accounts: UBIMET (for Morecast), and Ubimet AG (for 3 “wetter.tv” apps, meant for Austria, Germany and Switzerland). It’s worth noting that the website wetter.tv now offers weather reports for the entire planet, and it uses the weather icons from Morecast! And yet, the three wetter.tv apps still have that antiquated look, and there are small differences in the reported temperatures and the forecasts:

Strange thing, the wetter.tv apps show the current wind speed, whereas Morecast doesn’t do it–not even in the app!

As a matter of fact, there’s no reference whatsoever to Weather Underground and their over 250,000 PWS in any page about UBIMET, but in 2015 they said they “have access to over 28,000 weather stations around the planet, plus data from lightning detectors, satellites, radars, weather buoys and radiosondes. They developed their own model, called RACE, enabling high-resolution forecasts for highly accurate predictions in regions with complex topographies down to a resolution of 100 meters. UBIMET uses proprietary algorithms, high-quality data, and codes running on Intel Xeon processor-based clusters.” Note that the description for the 3 wetter.tv apps still references only “mehr als 17,000 WMO genormten Wetterstationen.” (I hope WMO is the German for PWS, because if it’s the the acronym for the World Meteorological Organization, it doesn’t have access to 17,000 weather stations, but only to stations for 2123 cities!) No clue as to whose those stations are, or whether they’re only the “official” ones belonging to the national weather administrations–and for the rest of the locations, probably calculated values on the maps?

[16]—Finally, the most outrageous “mainstream” weather service and app is AccuWeather (used as a data source by most OEMs), whose Superior Accuracy™ is one of the worst jokes ever, especially as its ridiculous MinuteCast® (“Rain starting in 23 minutes,” “A break in the rain in 1 minute”) is absolutely never correct! I didn’t check lately, but AccuWeather used to have one more quirk: their own app wasn’t in sync with their own website (even when just updated), i.e. they were showing different current conditions; 3rd-party apps using AccuWeather (probably via a different API or a different server) were showing a third set of current conditions–how pathetic is that? Here’s an informed testimonial:

Some similar love in Wikipedia: “AccuWeather … began increasing the range of their forecast from 15 days … to 90 days. These hyper-extended forecasts have been compared to actual results several times and shown to be misleading, inaccurate and sometimes worse than simple predictions based on National Weather Service averages over a 30-year period. It is generally accepted that the upper limit on how far one can reliably forecast is between one and two weeks, a limit based on both limits in observation systems and the chaotic nature of the atmosphere. An informal assessment conducted by Jason Samenow at The Washington Post asserted that AccuWeather’s forecasts at the 25-day range were often wrong by as many as ten degrees, no better than random chance and that the forecasts missed half of the fourteen days of rain that had occurred during the month of the assessment. … An assessment from the Post determined that the 45-day forecasts were not even able to predict trends accurately, and that, although the forecasts did not decrease in accuracy with time, the forecasts were so far off even in the short range to be useless.”

But wait! They released this communiqué: AccuWeather is Most Accurate Source of Weather Forecasts and Warnings in the World, Recognized in New Proof of Performance Results. To quote: “AccuWeather, the global leader in weather information and digital media, today announced additional verification information recognizing AccuWeather as the clear leader in Superior Accuracy™, keeping people safe and out of harm’s way around the world. … With the most complete global real-time and historical data, most robust database of forecast models, most advanced forecast engine globally, proprietary patents, and comprehensive validation results, AccuWeather is the most accurate weather company worldwide. … AccuWeather is recognized for its weather accuracy leadership in a new global report from ForecastWatch, a leading third-party weather forecast monitoring and assessment company. The study’s twelve-month evaluation names AccuWeather the most accurate source of overall temperature forecasts measured through mean absolute error and forecasts within three degrees. … ForecastWatch’s report of combined one- to five-day-out high and low temperature forecasts, ranking top providers of consumer weather forecasts worldwide, collected almost 11.7 million high and low temperature forecasts for 1,148 locations throughout the world for full-year 2015 and compared them with the actual high temperature, low temperature and overall temperature – the average of the high and low temperature – occurring each forecasted day. … Temperature is a main weather forecast data point for consumers and businesses that significantly impacts decisions across activities and industries, evaluated in the study.”

Here’s some nice pics supporting their claims:

Now, there are several idiocies in the above text, no matter what the tables say:

  • While forecasting temperature within 3°F (1.67°C) is nice, temperature is not the most important factor that “significantly impacts decisions across activities and industries”! What matters the most is to forecast heavy rain, snow and mixed precipitation, strong winds, T-storms, and extreme temperatures. When you’re missing most of the rain episodes, how can you still claim to be “the most accurate weather company around the world?
  • They forecasted some temperatures. To assess their precision, they had to compared to some measured data. What measured data? The problems with most weather apps is that there are no official weather stations for the respective locations, and for that reason many apps use PWS to display “current readings,” or they compute values based on satellite radar maps–both cases being imprecise and unreliable. Were those 1,148 locations selected to match exactly the existing official weather stations? There’s no mentioning of any such thing! So their claim of precision is bogus.

All in all, they’re bold and untrustworthy.

[17]—Qihoo’s 360 Weather also uses AccuWeather. The same stands for GO Weather Forecast & Widgets (aka GOWeatherEX). End of story. Or maybe not–see [41].

To be more specific: the Chinese apps are actually using Huafeng AccuWeather, a joint-venture. Beyond 360 Weather, plenty of Chinese have made similarly looking apps; one such example is Shenzhen UFO Technology Co., with two identical apps, this one and this one. (Never trust the Chinese!)

[18]A number of apps give you the choice of several data sources, but the choice is sometimes a false one. The usual sources:

  • YR.no (they have an app too, and it used to be acceptable, but with the new design both the widget and the app are less useful; also, it’s so stupid to see at 17:40 “Weather now 22:00-01:00,” when there are still more than 3 hours until 22:00 for which there’s no forecast; apps such as Detailed YR Weather Widget display the forecasts differently, yet they suck too) only gives forecasts outside Norway, not measured current conditions;
  • OpenWeatherMap and WorldWeatherOnline are crappy as hell;
  • and I’ve discussed Dark Sky (aka Forecast.io), Weather Underground, and Yahoo! Weather.

To add insult to injury, some apps ask you for an API key to use some of the data sources, because they don’t have one. Such multi-source apps include: Chronus; HD Widgets; Amber Weather (formerly EZ Weather) and its sister Amber Weather&Radar Free (formerly Amber Weather Elite); Transparent Clock & Weather; Weather & Clock Widget for Android; Meteogram Weather Forecast; and Sam Ruston’s acclaimed Weather Timeline, which has the insolence of trying to “Forecast up to years in the future” (“Open Time Machine”), while its various data sources can’t decide whether it’s sunny or it rains, or which temperature between -6°C and +2°C would be closer to the reality.

NOTE: The two “Amber” apps are now featuring exactly the same widgets and the same notifications, but the apps have different main screens. Even so, the weather data source configuration screen is identical in both apps!

[19]—Since many people seem to like the Yr app, here’s a screenshot-supported example of the complaint I made above:

At 11:12, I opened the app, and it showed me “Weather now 12:00 – 18:00,” which is preposterous! The 6 hours following the next full hour! Fortunately, I had a widget showing the conditions for “11:00 – 12:00,” and it didn’t get updated by the opening of the app. Nonetheless, the widget updated itself shortly after.

The 6-hr chunks are too large to be helpful, and there’s no way to display the previous chunk–although the current time might still be in that chunk!

Note that the maximum temperature (26°C) is shown for “18:00 – 24:00,” although several other apps pointed out that the maximum for the day (26-28°C) was to be expected in the interval 15:00 – 17:00. Obviously, if one could still expect 26°C at 18:00, by 24:00 the situation will be much changed, most likely to 19°C! Instead of 6-hr intervals, why aren’t they simply saying “here’s what you can expect at 6:00, 12:00, 18:00, and 24:00”? When someone sees a time frame, they expect the conditions to represent either the average, or the estimation for the median hour, e.g. for “18:00 – 24:00” that would be 21:00; in fact, what they display is for 18:00 at best. Morons.

But the worst thing about YR.no is that their website shows forecasts for every single hour of a day, whereas the apps only shows 3-hour chunks! Why?!?

UPDATE: On Sept. 15, 2017, NRK launched a beta version of their YR.no app! The beta installed from Google Play was version 5.2.2, as opposed to the official version 4.2.1. The widgets are pretty much the same, but the app itself bears some changes:

The forecasts are still only updated twice a day, which means the app can’t adapt to sudden changes that go against the initial forecast:

See [60] to learn more about the final release of this beta app.

[20]—A word on WorldWeatherOnline: while the current readings are in many cases for the closest airport (e.g. Stuttgart-Echterdingen), and the website asks for the ZIP, city name or IATA code, the forecasts are computed for the correct location. The official app sucks though, showing the maximum for each day instead of min/max, and the widgets are crappy too. One of the questionable Weather apps that use WWO (with decent graphics though) only updates the readings every 3 hours at best.

[21]—A few details on some apps with multiple data sources:

(1°) Chronus can nominally use 7 sources: AccuWeather (limited)OpenWeatherMap, YR.no, Weather Underground, Dark Sky, Weather.com (experimental), Yahoo! Weather. In practice, without having API keys one can only use 3-4 sources: YR.no, Weather.com (The Weather Channel), Yahoo! Weather–and the newly added AccuWeather. The most reliable is Weather.com (The Weather Channel), but unfortunately there’s a horrible bug somewhere, as the current readings seem to display the conditions for 1-2 hours ago for this data source only! I know that Weather.com is marked “experimental,” but this app is useless to me, despite the nice Pro widgets…

Let’s not forget that Chronus is merely a set of widgets, not an app! (Usually, it’s the other way around, i.e. apps without widgets.)

(2°) Weather Timeline is loved by everyone but me. It can use 5 sources: Weather Underground, Dark Sky, YR.no, OpenWeatherMap, WorldWeatherOnline. The last 3 sources are definitely useless to me. Switching between data sources is terribly annoying (too many settings), and choosing to use “feels like” replaces the current temperature instead of displaying both temperatures. Also, despite having so many widgets, I got sick of their stupid Material Design look. What pisses me off the most is the arrogant “Time Machine: Forecast up to years in the future”! And, of course, I can’t have two widgets using different data sources… But if Weather Underground or Dark Sky are good enough for you, this app might be a keeper. UPDATE: The absurd “Time Machine” has been removed in version 10.

Despite the ability to customize the icons, I still hate this app somehow:

(3°) Today Weather started as an app that could use Weather Underground or Dark Sky, but it now supports a total of 5 data sources, by adding Weather.com (The Weather Channel), AccuWeather (useless!) and YR.no. Since it displays the most stupid pictures I’ve ever seen (not from the actual location, and of course The Weather Channel does the same, but the pictures are less idiotic), I appreciate that the photos can now be disabled. Recent improvements added detailed widgets for the next hours or days, in addition to the stupid “now with stupid photos” ones.

But of course different data sources give different forecasts; Dark Sky vs. Weather.com vs. Weather Underground:

[22]—The Austrian developer pompously named Weather Services Group offers Weather Services and Weather Services PRO (3.29 €), and a choice of data providers between YR.no and OpenWeatherMap. The outcome is worse than the data providers themselves could offer, simply because the location is forced by the app to a stupid different one: the automatically detected Leonberg becomes Böblingen, and the manually set Leonberg becomes Tirschenreuth, Upper Palatinate (Oberpfalz), which is “the other Leonberg” (from BY instead of BW)!

[23]—Note on two meteogram (=meteorological diagram) apps that have corresponding widgets.

The most famous one is Meteogram / Meteogram Pro: while carefully designed as an app (and widget), I found it impossible to read on a smartphone that’s smaller than 6″. Also, the various forecast sources are highly divergent, so what should one choose? Even the source met.no, which should display the same forecasts as YR.no, doesn’t do that! (I’d rather select Weather Underground).

A less popular one (I wonder why) is Flowx (formerly WeatherBomb). The free edition can display the forecast for 7 days, the pro in-app purchase extends it to 10 days (but on a 5″ screen, 7 days is just fine). Data sources: USA NOAA GFS, Canada Weather GDPS. Normally NOAA’s GFS is probably the best choice (updated 4 times a day, resolution 100 km though; GDPS is updated 2 times a day, but it has a 25 km resolution). The problem with the widget is that there’s no legend, so it’s difficult to understand what’s depicted:

I guess the yellow means “clear sky” (“sunny” can’t apply during after sunset), the green line should be the wind, and the blue bump… the rain.

By customizing what’s shown (the “+” line in the app allows editing, not adding stuff!) and only asking for temperature, pressure in mmHg, rain and wind (with gusts), then by resizing the widget to almost the entire screen, here’s what I got:

The blue line is however the humidity, not the rain! To estimate the rain, one still has to squint at the top chart, so basically the widget should better be kept at 4×1 or 4×2.

Why can’t anyone make a really usable meteogram? (Yes, I know some other apps exist; even YR.no has meteograms in the app and on the website!)

[24]Exovoid.ch has a set of weather apps: the generic one, Weather XL Pro (translated e.g. Météo France XL PRO, Wetter Deutschland XL PRO, Meteo Italia XL PRO), and identical clones targeted to Austria, Belgium and Switzerland. They use Weather Underground with automatically selected PWS, e.g. ILEONBER18 for Leonberg, ISTUTTGA70 for Stuttgart, IBADENWU15 for Leinfelden-Echterdingen, because Weather Underground couldn’t care less that Stuttgart has an official weather station, and Leinfelden-Echterdingen is actually the airport!

[25]Don’t mistake the above apps for a different set of apps offered by a weather service that uses a more limited set of weather stations (e.g. for Leonberg it reads the current weather for Ludwigsburg, at 18 km), the Swiss MeteoNews. Apps using it are made by ID Mobile SA, and the universal one is Weather for the World, as the other 21 “Weather for <country_name>” are simply clones, conveniently translated for the local language (e.g. Weather for Germany also shows as Wetter für Deutschland). The unique widget shows the current conditions (or maybe a forecast for the hour?), but the app has the particularity that it only shows forecasts for “morning” and “afternoon” for each day. Also using MeteoNews.ch is the app LANDI Wetter, which has better widgets (and the small one is resizable), and for Leonberg reads the data from Stuttgart, not Ludwigsburg. Either way and for no matter what city, the long-term forecasts made by MeteoNews.ch are sometimes crazy: they differ from those made by most other sources, except for the one below.

[26]Speaking of forecasts instead of current readings, and of Swiss ones at that, meteoblue (which has a story and can use simultaneously e.g. 14 out of 26 weather models) severely lacks a widget and the app is uninspiring. Its forecasts can be very good, or on the contrary, quite crazy! (Meteoblue is nonetheless supposed to be more trustworthy than e.g. Dark Sky, ex-Forecast.io, see the corresponding section above. Except when it’s not!) If you look into the multimodel view, you’ll notice that some of the models are very divergent from others, so it all boils down to which model is trusted by any given weather app.

Here’s how divergent can those weather models be:

[27]Getting to “my” Germany and regardless of what weather data source is used, a number of local apps have to be mentioned.

First, wetter.com, an old friend I always revisit from time to time to notice the GUI improvements (I’m using this app since 2012). The current readings for my location are computed, not read (this is the difference between a small town and a large one), which can mean everything, from interpolated to forecasted. Too bad the large widget doesn’t display the wind for all the locations (I will discuss below such details). This app is not the best choice for a device with only 1 GB of RAM: the associated service, mguApi, takes about 32 MB of RAM, but the full app needs 120-160 MB.  If the weather fails to update in the widget, hitting the small gear then “Update Weather Data Now” should update the widget without fully opening the app.

Note that if for a certain day or part of the day the chances of rain are significant (e.g. 80-90%) but the expected precipitation minor (e.g. 0.5-1.5 mm or l/m²), the icon for the forecast won’t show any sign of rain, but the info about the precipitation is shown beside (risk of rain and expected precipitation):

The expected wind speeds are not shown for the entire day, but only in the details for a day, which include 4 time chunks: 6-11, 11-17, 17-23, and 23-5 (yeah, between 5 and 6 AM there’s nothing!):

More details about this app, at [30].

[28]—For a long time (and still today) I tend to use this as my main weather app: Weather & Radar Pro Ad-Free / WetterOnline Pro (2.99 €), by WetterOnline GmbH (they have a free edition too). Obviously, not always accurate, but decent enough. It tends to be pessimistic in forecasts when it comes to the minimum temperatures in winter and the maximum temperatures in summer; also about the chances of rain. Overall, trustworthy. The same information can be found online at WetterOnline.de.

Note that the app and its widgets have changed significantly in version 4! The round 2×2 (2×3 on some devices) widget can be resized to show more than the default 2 days and their maximum temperatures. Resizing it to the height of 2 rows and making it 3 or more columns in width (I keep it full width) gives something like this on my phone (on a tablet, the disc can be ovoidal for certain sizes of the widget and in landscape mode, so it might require different sizes on some devices):

While far from perfect, this app is extremely practical: not only it has widgets with forecasts for 3-4 days, but in the app, one can read on the same screen: hourly forecasts, daily forecasts, and min/max charts for 1 days!

And I particularly like this app during heatwaves, as it’s rather giving pessimistic forecasts!

Strange enough, this app doesn’t know the Brașov-Ghimbav weather station, yet it can display surprisingly accurate data for Brașov, sometimes more accurate than the official Brașov-Ghimbav station! I suppose it’s computed from weather maps, not retrieved from a local PWS. The web site lists Predeal as the nearest weather station for current readings, but the displayed data is different, so it’s not using Predeal instead of Brașov. A bit strange though.

[29]—Then, three apps that use the services of MeteoGroup, a private organization established in 1986 that owns a “private weather measurement network with approximately 1600 stations around the world”:

(1°) The official one, WeatherPro (2.99 €)–with a stripped-down WeatherPro Free edition–whose usability sucks and which declined in quality, not to mention that they now want a subscription even to let you change the icon theme, but somehow it’s well-regarded. The same contents is available on wetter24.de.

NOTE: Don’t mistake MeteoGroup’s site wetter24.de with wetter.com GmbH’s site weather24.com. Beside wetter.com, wetter.com GmbH also owns www.weather24.comwww.eltiempo24.es and www.tiempo.es, while MeteoGroup also owns www.wetteralarm.at, www.meteo-info.be, www.vejrcentral.dk, www.saa-varoitukset.fi, www.vigilance-meteo.fr, www.unwetterzentrale.de, www.severe-weather-ireland.com, www.meteo-allerta.it, www.meteocentrale.li, www.meteozentral.lu, www.noodweercentrale.nl, www.vaer-sentral.no, www.centrometeo.pt, www.alertas-tiempo.es, www.vader-alarm.se, www.meteocentrale.ch, www.severe-weather-centre.co.uk.

(2°) wetter.info, which is promoted by Telekom.de and preferred by some newspapers and media outlets, and which sometimes gives differences of up to 4°C between locations are 3-4 km away (70839 Gerlingen vs. 71229 Leonberg), or which can have such a lack of new data that it can display the same temperature as 3-4 hours ago; their website reveals that for Leonberg they use the readings from MeteoGroup’s Renningen-Ihinger Hof weather station (some 8 km away as the crow flies), also for forecasts, and that for Gerlingen they use the readings from MeteoGroup’s Stuttgart-Wilhelma weather station (some 11 km away from Gerlingen), which explains the discrepancies: while the distance Leonberg-Gerlingen is 3-4 km, the distance between the weather stations used for readings is 25 km. The same contents is available on wetter.info.

(3°) RTL’s Wetter.de, mostly neglected, despite some interesting features, such as the widget showing the chances of rain for the next hour and the forecast for two of the following periods of the day that follow (Mittag, Abend, Nacht, Morgen), and extremely detailed hourly forecasts in the app; RTL’s app can however show slightly different data than the official MeteoGroup, especially for small towns, its “Gefühlt” (Feels like) is definitely different in this app, it’s a bit slow and heavy, and it inherits from MeteoGroup the flaw of failing to provide recent current temperatures for some small towns, making it even worse, with even longer delays–say, current temperatures don’t change for 2-3 hours (the website actually admits it, e.g. by mentioning that the temperature displayed at 20:30 was measured at 19:00). The same info is available on wetter.de.

Too bad the large widgets don’t include more information; e.g. not only the maxima, but also the minima for the next days:

All things considered, from the apps based on MeteoGroup’s network, the one recommended by Telekom.de is the worst!

[30]—A few more issues about wetter.com. I still can’t make my mind about it. They do have their own network of affiliated PWS, yet it’s by no mean clear how the entire thing works, and they also read the official national (DWD in Germany) weather stations, plus the airports. One of the strange things: some current readings are “measured” (e.g. for Stuttgart), others are “calculated” (e.g. Leonberg, Gerlingen). Even for those that are calculated (does it mean they’re interpolated on a map?), some are including wind information (e.g. Gerlingen), some others don’t (e.g. Leonberg, despite being larger than Gerlingen, and only 3-4 km away)–the wind info is only shown in the app and in the 4×2 widget.

The current readings are usually different even for small towns that are 4-10 km away, i.e. the following locations are having different readings 71229 Leonberg, 70839 Gerlingen, 71254 Ditzingen*, 71272 Renningen, 72760 Reutlingen; and different from 70173 Stuttgart-Mitte* (which seems to be reading Wetterstation Stuttgart-Schnarrenberg 70376) and 70771 Leinfelden-Echterdingen* (Flughafen Stuttgart STR/EDDS). The names with asterisk are having the temperatures reported with decimals! OK, so wetter.com strives to provide localized readings and forecasts. Most of the times, for Leonberg (Altstadt), the readings for the neighboring Gerlingen seem to be more accurate: are the readings for Leonberg computed or retrieved for its south neighborhood of Eltingen? Either way, it’s sometimes funny to see how the current conditions can toggle every 15-30 minutes between “here it’s raining, 4 km away it’s not” and “here it’s not raining, 4 km away it rains,” while other apps don’t bother to budge.

On their website, for all of the above small towns, the following are listed: “Unsere Wetterstation aus 71116 Gärtringen (17.2 km from Ditzingen, 18.8 km from Leonberg, 19.8 km from Gerlingen); Unsere Wetterstation aus 74379 Ingersheim (23.9 km from Ditzingen, 22.5 km from Leonberg, 21.3 km from Gerlingen).” OK, so what are the sources for those current readings? Ditzingen is “measured” (not “computed”), so there must be a PWS there!

But sometimes I feel that the app is cheating. For instance, both Leonberg and Gerlingen are “computed,” yet only Gerlingen gets the current wind speed! By “pure coincidence”–what else?–the direction and speed are always identical to those measured for Stuttgart (Mitte). Probably Gerlingen is considered “close enough”–but not Leonberg. Otherwise, the current sky condition can differ between Gerlingen and Stuttgart (Mitte), only the wind is “borrowed”…

Based on what I’ve noticed about the apps based on MeteoGroup’s network of weather stations, it could have been that wetter.com too is reading Renningen-Ihinger Hof (which doesn’t provide wind measurements) for Leonberg and a different station for Gerlingen. As it happens, Renningen-Ihinger Hof is also used for Leonberg by DWD’s own WarnWetter app. However, very often the temperature shown for “Klimastation Renningen-Ihinger Hof” is slightly different, so no, this app doesn’t use it for Leonberg.

A last hint on using wetter.com for locations in Romania: search for “Wetterstation Bucuresti Baneasa” instead of Bucharest and for “Wetterstation Brasov-Ghimbav” instead of Brașov! Even so though, the app won’t always display measured current values, but very often calculated values even for locations that are a Wetterstation or an airport (Flughafen), which is a shame! Alternatively, the measured values can be older than 1 hour.

Another queerness (shared by other apps though). Suppose the hourly forecast reads as follows: 17:00 = 25 °C; 18:00 = 24 °C; 19:00 = 23 °C; 20:00 = 22 °C; etc. Suppose there’s 17:30 and the measured or calculated temperature is already down to 22 °C: do you expect them to adjust accordingly the hourly forecast? NO WAY! The app will still pretend that after 30 minutes the temperature will somehow be as high as to be able to decrease according to the forecast, i.e. 22 °C → 24 °C → 23 °C → 22 °C, despite the day being on the decreasing slope of the temperature!

Oh, I forgot to mention that this app too tries to downplay the heatwaves, e.g. by forecasting 30 °C instead of 33 °C. This kind of downplaying is however also specific to DWD.

And a funny bug about wetter.com: the large widget and the app don’t show the same sunrise and sunset times! E.g. for 71229 Leonberg on Sept. 26, 2017: ↑07:17 ↓19:13 in the widget, ↑07:19 ↓19:19 in the app and on the website.

UPDATE, OCT. 2017: In version 2.21, released on Oct. 4, 2017, wetter.com finally got a widget able to show more than the current conditions! They replaced all the widgets (but the webcam one) with a resizable 4×2 widget that can be made to show more or less data, depending on the size:

There is still a problem with this widget and this app: it shows it’s raining, the app has already forecasted several episodes of rain for the day, and yet the icon for “Today” only shows an unperturbed sun! This is unacceptable. The wind warning also doesn’t show on the icon for today, although in the app there is a wind cone on the icon; notice the 90% chance of rain and the 5.9 mm of expected rain–why the heck is the icon for the day not showing any kind of suggestion of a rain?!

No matter how much they fiddle with it, this app will never get the things right. I’m disappointed.

[31]—Since I mentioned DWD’s WarnWetter app, and its exclusive use of the official weather stations (no PWS whatsoever)… The app itself is quite fine, and for the automatically detected Leonberg it displays as the “belonging” weather station Renningen-Ihingerhof. There is however a bug with the widget: when set to GPS detection, it detects Leonberg, but it selects Engelberg (Winterbach) as the weather station; only when the location is manually set to Leonberg it uses Renningen-Ihingerhof. This weird bug might be due to the fact that there is an Engelbergturm in Leonberg. Engelberg (part of Winterbach) is however some 33 km away, while Ihingerhof (part of Renningen) is only 8 km away.

I forget to mention that DWD’s widget has a bug: I had to resize it to the height of 2.5 rows, because at 2 rows it wouldn’t show the 4-day forecast! (Thankfully Nova Launcher allows that, otherwise I’d have needed to make it 3 rows high.) YMMV: maybe it depends on the screen size and the PPI, the font scaling, etc. I expect the widget to work better on 5.5″+ screens.

On Android 7+ there’s a thing called “display size” (not “font size”!) that allows a sort of “zoom out” for the screen, so that more contents would fit (in contrast, “font size” only changes the DPI for the text). Here’s how the DWD widget shows right after it’s dropped on the screen of my 5″ Moto G5 with the default settings…

…and here’s how it looks like with changed metrics–“display size” set on Small and “font size” set on Large, to compensate the zoom-out:

[32]—To illustrate how the PWS readings for Gerlingen are usually more reliable than those for Leonberg, especially for the upper part of Leonberg (Altstadt/Marktplatz, not Eltingen), it’s perhaps worth noting that most of the times Weather Underground is defaulting–for my location, close to Marktplatz–to PWS such as IBADENWR213IGERLING9 or IGERLING56 (the latter hilariously labeled Leonberg) instead of the closer ILEONBER62ILEONBER10, or ILEONBER3. There must be a reason for that…

Oh, and I noticed that in Yahoo! Weather too the readings for Gerlingen are closer to the reality than those for Leonberg. It’s almost as if it would use the same data sources as wetter.com.

[33]A special clarification regarding 3 apps with very similar names.

The 1st one, El Tiempo 14 días–translated to Weather 14 Days, Wetter 14 Tage, Météo 14 jours–belongs to Meteored and corresponds to the websites tiempo.com, tempo.pt, daswetter.com, ilmeteo.net, yourweather.co.uk, tameteo.com, theweather.com (there’s a Pro version too for 1.99 €, which was discounted at 0.10 € for a while). I couldn’t identify the weather data source, but the app is forecast-oriented, and I strongly believe that the current conditions are simply the forecast for the hour, and never measured values. This would explain why the app seems at times so accurate and some other times so very wrong (not only for my location; I’ve read many reviews in several languages). This app’s forte is that it has wind forecasts for each hour and every day, however the shown value is the maximum (gusts) one, and it’s highly exaggerated even so (the average wind speed is hidden in the details for each hour). I’d also have preferred to see the probability of rain, not only the expected amount of rain (0 mm of rain expected, but 0% or 35% chances of rain? 1 mm of rain expected, but 40% or 85% chances of rain?).

Note that this app is only too eager to use the icon “Cloudy intervals” for the entire day or for each and every hour of a day, even if the day still has a decent amount of sun and clear skies; it will also show a day as rainy even if the rain is going to minimal (if any).

UPDATE: Mid-2017, they said they were in the process of switching their forecast model to ECMWF (European Centre for Medium-Range Weather Forecasts), based in Reading, England, UK. You’ll find a lot of furious reviews on Google Play, most of them from February to April 2017, complaining about the accuracy of the forecasts. To be honest, YR.no‘s forecasts can be as bad as this app’s ones, but Weather 14 Days still has some extra faults: it doesn’t know most cities of Japan, it doesn’t know of Cyprus, and even some locations in Scotland are too difficult for it. Also, whereas it’s not a bad looking app, I find it looked much better back in 2014!

Extra info about forecast models, from WetterZentrale.de:

  • GFS is the global weather forecast model of the US weather service run at an internal resolution of 28 km. It offers a plethora of parameters for the next 15 days. Updated 4 times a day up to 384 hours ahead. The runs for the 0, 6, 12 and 18Z runs are usually coming in from 3:30, 9:30, 15:30 and 21:30 UTC, respectively.
  • ECMWF: The European Centre for Medium-Range Weather Forecasts (ECMWF) creates forecasts for the upcoming 15 days and is a global leader in forecast skill. However, it offers only a small number of parameters for free. The 00Z and 12Z runs are coming in twice daily between 6 and 7 UTC and 18 and 19 UTC.
  • WRF (Weather Research and Forecasting model) is a freely available weather forecast model. WRF version 3.9 is run on the WZ servers and offers 4 forecasts per day up to 72 hours. It is currently run with a horizontal grid spacing of 5 km and uses the WRF single-moment 3 class microphysics scheme, the YSU PBL scheme and the Dudhia radiation scheme.
  • ICON is the global weather forecast of the German weather service (DWD) with a grid spacing of approx. 13 km globally (Europe nest: 6 km). Model output is available up to 180 hours ahead for the 0Z and 12Z runs, and up to 120 hours for the 6Z and 18Z runs.
  • HIRLAM is a regional model used by many weather services. WZ offers the Dutch (KNMI) and Finnish (FMI) version. It covers Europe with 10 km resolution and is calculated 4 times per day with 48 h forecasts.
  • ARPEGE is the global forecast model of the French weather service (Meteo France). It is runs with a maximum resolution of approx. 7 km in Europe and mean global grid spacing of 15 km. WZ offers forecasts up to 102 hours.
  • GEM is the global forecast model of the Canadian weather service and computes 10 day forecasts.

From meteolo.com:

  • GFS: The American GFS (Global Forecast System) model run in two parts: the first part (higher resolution) goes out to 192 hours with charts for every 6th hour. The second part (lower resolution) runs from 204 to 384 hours with charts for every 12th hour. In other words, first part goes to 8 days and second part goes to 16 days. This numerical model is run four times a day, using 0, 6, 12 and 18 UTC data, and charts update begins at 3:30, 9:30, 15:30 and 21:30 UTC. First part resolution charts is about 100km (1º), and second part is about 250km (2.5º). This model is run by National Oceanic and Atmospheric Administration (NOAA).
  • ECMWF is best known for this global operational forecast model, been officially known as the “Integrated Forecast System (IFS)” but usually known informally as the “ECMWF” (European Centre for Medium-Range Weather Forecasts). The IFS is a global model that runs every 12 hours and its output runs out to 10 days (240 hours) in 24 hours intervals. Original IFS resolution is about 50km (0.5º), but charts resolution is interpolated to 100km (1º).This numerical model is run two times a day, using 0 and 12 UTC data; charts update begins at 6:00 and 18:00 UTC.
  • UKMO, also known as UKMET (United Kingdom Met Office), is the global model product of Unified Model. The UKMO model runs every 12 hours and its output runs out to 3 days (72 hours) in 6 hours interval (6h to 48h) and 12 hours interval (48h to 72h), with a resolution about 125km (1.25º). This numerical model is run two times a day, using 0 and 12 UTC data; charts update begins at 3:50 and 15:50 UTC.

From Windy:

  • ECMWF (European Centre for Medium-Range Weather Forecasts) is sometimes informally known in the USA as the “European model”. Resolution is 0.1° (~9km). The forecast output is produced every three hours for first 144 hours, six hourly through day 10.
  • GFS (Global Forecast System) – the model is operated by the United States’ National Weather Service (NWS). Resolution is ~13 km. The forecast output is produced every hour for the first 120 hours, three hourly through day 10.

Back to our app, here’s El Tiempo 14 días/Weather 14 Days/Wetter 14 Tage shown below WetterOnline/Weather & Radar Pro:

And here’s its biggest widget:

The 2nd one, El Tiempo por eltiempo.es, belongs to Pelmorex Media Inc., a network that has also a different app–The Weather Network / MétéoMédia–and the websites ElTiempo.es, TheWeatherNetwork.com, MeteoMedia.com, WetterPlus.de, Clima.com.

The  app made for Spain, El Tiempo, seems to be 100% forecast-driven, which suggests that this might also be the case for the “normal” app The Weather Network / MétéoMédia.  In the past, the app used to give the same readings for Leonberg and the Stuttgart airport, but now the app and the website are showing different values for Leonberg, for Stuttgart and for the airport. The values for Leonberg are rather wrong though. It’s still difficult to find the correct location at times, e.g. there is “Bucharest” for the city and “Bucuresti Otopeni” for the airport.

The 3rd one doesn’t even exist for Android, but because of a possible confusion, here it is: tiempo.es is the site that belongs to wetter.com GmbH (despite being managed by Olivares Media SL), and the iOS app, tiempo.es Universal, has the exact same UI as the iOS wetter.com app; the Germans just wanted to have two identical iOS apps (with different logos and names) for the Spanish market!

[34]—As it happens, I discovered some ways ForecaWeather is better than MSN Weather.

Remember what I said about the apps that present the idiocy of showing for any given day a pair of temperatures that are not the minimum and maximum temperatures for that day, but the maximum for that day and the minimum temperature for the night that ends in the next day’s morning? In simpler words, it’s as if a day would start at 7 AM and would go through 7 AM of the next day! I just noticed that the list of such apps doesn’t only include The Weather Channel, Google News & Weather, and Weather Wiz, but also… MSN Weather!

As ForecaWeather and MSN Weather use the same data source (FORECA), it’s easy to compare the two widgets with regards to the max/min for each day:

Pay attention how the minimum of -1° for SAT morning (between 2 and 7 AM) moves in the case of MSN Weather to the day of… FRI. The minimum temperature for the current day, which is FRI, is nowhere to be found in that widget (it was +1°C, as per FORECA). Similarly, FORECA says that the following MON the temperatures will range between +1° and +13° for the entire period of 24 hours (0:00 to 23:59:59 military time). MSN prefers to put that +1° information attached to the day of SUN, and MON shows the minimum temperature for the morning of TUE! I simply cannot use such apps designed for morons! Any day starts at 0 AM and ends at 0 AM the next day, full stop! I don’t need “the minimum for the night that ends at 6-7 AM the next morning,” you assholes!

Incidentally, Yahoo! Weather shows the max/min correctly, for the actual day (0-24, not 7-to-7).

Some facts about ForecaWeather. It indeed cannot show measured temperatures for my location, but for an official station some 18 km away. Still, the forecasts are pretty good, and for my exact location (computed from maps, mind you). If the widget shows the wind and the “feels like” for a somewhat distant location, the information is still very useful. Moreover, in the app, in the default “Day Forecast” view, swiping to left allows to see–for each of the next 10 days–hourly forecasts with: weather icon, temperature, chances of rain, wind (average speed, not gusts), “feels like” and humidity. Unparalleled richness of info… except there’s nothing about the forecasted maximum (gust) wind speed! For such info, Weather & Radar / WetterOnline (even the free edition) or wetter.com (which can also show DWD’s weather warnings) are better choices.

While Yahoo! Weather (the dark screen above) doesn’t specify the maximum wind speed, but only the average (the forecast for the day below the “Details” section, and the current wind below the map), its hourly forecasts use a specific icon for the hours with a particularly strong wind–which makes my choice for this pair of weather apps even more justified.

I said that if the weather station for current readings doesn’t come with new data, this app would serve old data, and that sometimes, somehow, MSN Weather manages to work around this, possibly by switching to a different station. Still, at least ForecaWeather shows the exact weather station and the time of the measurement, which is more honest than in most other apps! On the other hand, even when the weather station does serve new data, MSN can have a 30-minute delay in retrieving it, no matter it says it just updated the data–it actually got the same old data from the server! I’d rather stick to FORECA’s official app…

…which is not without bugs. Its Windows 10 (ModernUI) app is much less usable, and it defaults to the wrong station, which cannot be changed. For Leonberg, instead of “Observed at Stuttgart / Schnarrenberg,” (which is reasonably close to Stuttgart-Mitte) the Windows app reads “Observed at Stuttgart-Echterdingen,” which is actually the airport at Leinfeld-Echterdingen (Flughafen Stuttgart). Unacceptable, so I’l stick to the Android app, which is actually newer… and has two bugs on its own. First, when you add the widget, on some devices it will get stuck with “Waiting for data…”–until you’ll resize it back and forth, which will repaint it. (The 4×2 widget adapts and shows fewer information when resized as 4×1, 3×3, 2×2, 1×2 or even 1×1, just try it!) Second, despite in theory being able to display the units as I wanted them to be (km/h and mmHg), the third metric option (see below) actually displays m/s and mmHg instead of km/h and mmHg. Can anyone really think of a speed in m/s?!

Oh, the nice thing with FORECA is that their “feels like” is most of the time more pessimistic than MSN’s; I guess it computes the “wind chill” effect based not only on wind’s speed, but also on the humidity. More realistic, I’d say.

But here’s why I can’t use this app anymore: in a screenshot above, the full-width (4×2) widget was showing a good deal of info for the current conditions, and 4 days of forecasts. That was on a 5.5″ screen. Simulating a tablet in Nox App Player, I learned that the widget can show even more details (pressure, etc.) and 6 days of forecasts! On my new 5″ Moto G5 though… shock and awe:

Note that the change of metrics mentioned at [31] for DWD doesn’t work here–the range is not large enough. This widget is clearly designed for tablets.

And yet, FORECA is not bad if you only use their website. It would absolutely show you the weather station used for readings, which is crucial. While I’m not happy that for Leonberg (BW) it cannot use the weather station at Renningen-Ihingerhof, for most locations it should be quite good. Funny facts for the readers in Romania: for Bucharest it uses by default Bucaresti (sic!) Filaret, and the website allows you to switch to Bucuresti-Imh (which is ANM, ex-INMH Băneasa) or to Bucuresti Otopeni (the OTP/LROP airport); for Brașov it will use Brasov-Ghimbav; and it knows of Sfantu Gheorghe Govasna (sic!).

[35]—I only mentioned that Gismeteo is among the apps that use the Stuttgart Airport readings for Stuttgart and a number of towns around, which is rather unpleasant and worse than FORECA’s use of Stuttgart-Schnarrenberg. Still, Gismeteo has its forecast charts computed for the actual location, not for the airport, so that the 4×2 hourly widget can prove itself useful enough, if one doesn’t want to know with accuracy the current conditions. Too bad I’m not satisfied with the forecasts: the app is too eager to put the sunny icon on the entire day when all the other apps forecast “partly cloudy,” or to put the moderate rain on a day for which no rain is expected at any hour (the second fault is shared though with Yahoo! Weather). And that they never heard of DWD’s weather station Stuttgart-Schnarrenberg. In the “Nearby airports and weather stations” tab, Gismeteo lists 2 airports and 7 DWD stations and shows their readings: Stuttgart Airport 19 km, Reutlingen 35 km, Heilbronn 40 km, Horb am Neckar 46 km, Sinsheim 50 km, Karlsruhe 51 km, Freundenstadt 57 km, Karlsruhe/Baden-Baden Airport 68 km, Heidelberg 71 km. (The website would even show the temperatures with decimals, something the app can’t do.)

[36]—How about the weather apps that claim to be so smart as to predict that “Rain will start in 19 minutes and will go on for 48 minutes”? Beside AccuWeather, many other apps include such immediate forecasts: The Weather Channel, Weather Underground, and more (in US/UK, Dark Sky and other apps using them for a data source). Some providers even offer dedicated companion apps: Storm Radar (by The Weather Channel); Storm (by Weather Underground); RainToday (by MeteoGroup). Needless to say, the last one is quite ridiculous, and not only when it claims to know when the rain will start or stop, but also when it says no rain is to be expected, yet the icon shows otherwise (but “it doesn’t compute”); at least, it usually knows when it rains:

Some other apps, such as Weather & Radar Pro / WetterOnline Pro, only offer approximate estimations for the next 15-30-45-60-75-90 minutes. That’s more reasonable. (The Weather Channel also can be reasonable when it only mentions “chances of” rain or thunderstorm at “about” a certain time.)

A third category are the apps that display the official weather warnings from the national weather service or from Meteoalarm.eu. For Germany, it’s the case of DWD’s own WarnWetter app and of wetter.com, both displaying the full text of the notification, then MeteoGroup’s WeatherPro, displaying a succinct version; The Weather 14 Days displays all the warnings for the region, even those not applying to the exact location; The Weather Channel also takes into account the official warnings.

While the official warnings can prove wrong or overcautious, they’re usually also based on observations, not just on mathematical models; but the “precise” warnings based on computations alone are ridiculous. This is why I couldn’t trust apps that claim to know exactly when the rain is going to start and how much it’s going to last. Pure crapola.

[37]AD HOC TEST, AUG. 2017: I’ve been disappointed more than once by all the weather apps, but let’s make a short test, some 45 minutes after a rain started. Everyone predicted this rain, it was in all of the forecasts, and it was supposed to start not earlier than 5 PM and not later than 7 PM. It actually started around 5:30 PM, and the result was that the temperature fell suddenly from 26°C to 19-20°C even on my window thermometer—the wind was that cold! I’ll use as a reference the official DWD readings for Renningen-Ihingerhof (4 km away) made at 6 PM, simply because they’re reading 19.9°C at 6 PM, as my own thermometer managed to fell slightly under 20°C. Note that there was an official warning for strong winds (gusts up to 80 km/h), but the local station doesn’t measure it.

Here’s what the apps were reading at 6:30 PM; many were still thinking that the heat of the day was still with us:

  • The Weather Channel: 21°C (Android), 20°C (Windows). This app has some weird bug: the automatically detected location shows a different temperature than the manually entered one (typically by 1°C); this bug is also present in Google News & Weather. 15 minutes later, 19°C (both OS).
  • Dark Sky: 22°C (but this is a forecast). 15 minutes later, 21°C. Wind: 10 km/h.
  • Forecaster: 23°C (as usual, lagging behind the official app). 15 minutes later, 22°C. Wind: 10 km/h.
  • Weather & Radar Pro: 23°C (disappointing). 15 minutes later, 20°C. At least 45 minutes late!
  • wetter.com: 20°C (quite a surprise, but “calculated” as opposed to “measured”), forecasted winds 24-52 km/h.
  • WeatherPro (MeteoGroup): 24°C (and 16 km/h wind) in the app and the screen widget, and 20°C (and 39 km/h wind) in the notification area widget. No manual update would persuade the app to sync the two readings for the next half hour!
  • MSN/Foreca: 22°C (and 29 km/h wind), but measured in Stuttgart/Scharrenberg (DWD also measured 22°C in Stuttgart/Scharrenberg).
  • Weather 14 Days: 20°C (another surprise, but it might be a forecast), forecasted winds 12-43 km/h.
  • Meteoblue: 26°C (the most ridiculous of them all!), wind 10 km/h

Well, Deutscher Wetterdienst (DWD) was also a bit stupid when their Stuttgart Airport station was measuring 22°C, and METAR more appropriately said 20°C.

By 7:30 PM most apps were reading less than 20°C, while Renningen-Ihingerhof measured 16.6°C at 7 PM:

  • 22°C: Forecaster (no change after 7:30 PM), MSN/Foreca (reading Stuttgart/Scharrenberg)
  • 21°C: Dark Sky (20°C after 7:30 PM)
  • 19°C: WeatherPro (MeteoGroup), Weather 14 Days, WorldWeatherOnline (reading Stuttgart Airport though), OpenWeatherMap, YR.no (being a forecast though)–the last three in Weather Timeline
  • 17°C: The Weather Channel (and Google), Weather & Radar Pro
  • 18°C: wetter.com, Meteoblue (finally!)
  • 16°C: Weather Underground (the default PWS, whatever it was, in Weather Timeline)

I had 18°C at my window, so I suspect the right reading should have been in the range 17-18°C, but not quite 16°C, and definitely not 20°C.

The clouds were supposed to dissipate completely by the next morning, with most apps forecasting no rain after 2…8 AM (no two sources would agree on when the chance of rain would become negligible). DWD said it won’t rain after 8 AM, and the last drop of rain fell around 7 AM. Overall, given the high volatility of the conditions (changing wind), most apps performed reasonably well.

I monitored the readings for the next 48 hours, and here’s a summary of my conclusions:

  1. The current temperatures can unexpectedly and ridiculously go astray with YR.no, Meteoblue, Weather 14 Days (obviously, not counting the apps that are usually wrong for my location, e.g. WorldWeatherOnline, OpenWeatherMap, MSN/FORECA); nonetheless, Weather 14 Days has rather decent forecasts (or does it?).
  2. The current temperatures and conditions can sometimes be extremely accurate with The Weather Channel (but not always, alas!), and the warnings and forecasts are typically useful.
  3. Inconsistent quality, ranging between “excellent” and “WTF” for Weather & Radar Pro (which makes me angry)! This app still has the best weather radar of them all.
  4. I still don’t have a clear opinion on Dark Sky (I’ll still keep an eye on it).

Already having trustworthy forecasts for 48 hours is a challenge, but how about long-term forecasts? Consider this chart, where I marked the websites instead of the corresponding apps:

How about the risk of raining for beyond the next 72 hours?

I couldn’t use YR.no and Weather 14 Days (daswetter.com) simply because they only show the amount of rain, as if the chances of rain could only be either 100% or 0%. Strange thing, Weather Underground didn’t match The Weather Channel (weather.com), despite being the same company now.

Some such sources would update their forecast–sometimes with radical changes–daily or more than once per day (especially wetteronline.de, which is Weather & Radar Pro), but this is the initial forecast made by eight different sources. I couldn’t be bothered to check the actual temperatures, and it wouldn’t be relevant, as the forecast changes as the days go by.

And then… surprise: Deutscher Wetterdienst (DWD) revised on Tuesday their medium-term forecast to decide that there’s going to be no rain on Thursday and Friday! Here’s the 3 successive forecasts, each check being made at about 7 hours after the previous one (at about 0:30, 7:30 and 14:30):

To simplify:

  • Forecast 1: After midnight: chances of small rain on Thursday afternoon, definite chances of rain on Friday, particularly in the afternoon (but also in the early morning).
  • Forecast 2: In the morning: no rain whatsoever on Thursday, modest chances of rain on Friday (in the afternoon).
  • Forecast 3: In the afternoon: no more rain until the weekend, yay!

How were other weather sources reacting to these changes in the weather models?

A couple of hours after the 3rd forecast:

  • The Weather Channel and YR.no basically had the same forecast as DWD. Going beyond the icons though, and reading the details in Google News & WeatherThe Weather Channel still gave 18% chances of rain for Thursday morning ~4 AM and 13% ~4 PM, and 17% for Friday ~4 PM. About 8 hours later, the chance of rain for Friday afternoon raised to 30%, and Google News & Weather changed the icon for the day to rain–and this is when I uninstalled it!
  • Weather 14 Days: 0.3 mm of rain on Friday around 17:00. As this app is only to eager to show the clouds or the rain in a day’s icon, the icon for Friday meant rain. Hours later, it dropped the rain, but … the next day, after 11 AM, the 0.3 mm of rain on Friday showed up again, and the day icon was showing rain too!
  • Weather Underground read in Weather Timeline: the icons didn’t show any rain, but for both Thursday and Friday, “A stray shower or thunderstorm is possible.”
  • Weather & Radar Pro: still showing 60% chances of light rain for Friday. They refused to update the forecast even 10 hours later! Eventually, by next morning, Friday was shown as rain-free.
  • Dark Sky: chances of rain for Friday 17%, with light rain (13%) ~4…6 AM. The same data source was still showing the older forecast in Weather Timeline, i.e. Thursday “Rain starting in the afternoon” (37%), Friday “Light rain in the morning” (37%). Strange thing, they never thought it’s going to rain on Friday afternoon!

Here’s the Weather 14 Days sudden fuck-up, when the 0.3 mm of rain screw the icon for Friday just as Weather & Radar Pro stopped showing any rain:

Prepare to be puzzled: the next day, NOAA’s GFS model, as shown in Flowx, still shown meager chances of rain for Friday (and more definite chances for Sunday towards Monday):

[38]—All things considered, here’s a summary of my complaints regarding the apps and the data sources I’ve used the most, and after having used them to monitor Leonberg (BW), Bucharest, Brașov and Sibiu:

  • The Weather Channel is generally a decent source, but:
    (1) to them, the current day ends next morning at 6 AM, which is ridiculous;
    (2) the official app has some bugs and (most important!) no widget to show more than the current conditions;
    (3) using it as a data source in other apps tends to display other bugs: either delayed sync, or the inability between 12 AM and 6 AM to show the maximum for the current day because, in order to show the minimum for the current morning, the app is querying the data for the previous day (this happens in Today Weather); the number of apps able to use weather.com as a data source is still very limited (and Weather Wiz is such a fuck-up, see [11]).
  • Weather Underground cannot be trusted:
    (1) because it uses more or less random PWS of doubtful reliability;
    (2) because changing the default PWS for a location is not obvious in the official app, and impossible in 3rd-party apps;
    (3) because the reported temperature is typically higher than normal on sunny days (the PWS is not hosted in the proper conditions), and because the PWS might have stopped reporting for 2-3 hours before Weather Underground decides to switch to another one.
  • MeteoGroup’s WeatherPro and all the apps using the same data source:
    (1) are unreliable even for Germany, despite being a German company, simply because the current conditions can sometimes be 2+ hours older;
    (2) have widgets that annoy me in a way or another, or have other usability issues. Dismissed.
  • wetter.com, despite being nice:
    (1) still lacks a widget to show more than the current conditions;
    (2) has questionable forecasts;
    (3) forecasts do not show any sign of rain in the icons even for pretty decent chances of rain, if the amount of rain is small;
    (4) displays current conditions that frequently are more than 2 hours older for some locations outside Germany or more than 1 hour older for German locations;
    (5) for “Wetterstation” or “Klimastation” locations you’ll often find “calculated” instead of “measured” temperatures and conditions. Outrageous. Dismissed.
  • Weather & Radar Pro (WetterOnline):
    (1) has the best weather radar of all;
    (2) has very good widgets;
    (3) can show expected strong winds in a day’s icon, despite not being able to show the wind for current conditions;
    (4) forecasts only show rain in the icons for chances over 50%;
    (5) has decent current conditions, but not always correct;
    (6) has jumpy and extreme forecasts, e.g. changing to often or updating too late, forecasting too high temperatures, etc. The app still deserves some attention, no matter where in the world you are.
  • Dark Sky (ex-Forecast.io)…
    (1) is not a reliable source for outside US and UK;
    (2) it uses the rain icon for the entire day even for light rain chances of 11%;
  • Data sources and apps dismissed altogether:
    (1) YR.no, for their approach (see [19]), and for the lack of reliability for my location;
    (2) FORECA and MSN Weather, for design bugs (see [02] and [34]), and because for my location the available weather stations for current readings are too far away;
    (3) Yahoo! Weather: not very reliable, questionable forecasts, and no widget to show several days;
    (4) Meteoblue.ch: ridiculous current conditions on occasions (forecasted, not measured!), despite the multi-model approach, and no widget at all.
  • Out of the question:
    (1) AccuWeather;
    (2) OpenWeatherMap;
    (3) WorldWeatherOnline.
  • Weather 14 Days is a special case, because beyond the fact that it’s now using ECMWF, there’s no indication on whether the current conditions are measured or computed from maps!
    (1) while it’s generally mostly decent, when it displays clearly wrong current conditions it does so very much the same way YR.no is doing when it’s in the wrong, which suggests it only displays forecasts;
    (2) the icon for a day will show rain even if the expected amount of rain is insignificant;
    (3) most days will have an icon with “Cloudy intervals” even on days with plenty of sun (but some clouds now and then);
    (4) the app can’t show the chance of rain, but only the amount of rain (if any);
    (5) it’s one of the few apps to show wind data in the app and in the widgets;
    (6) has rather good widgets. So it deserves some attention, despite its hit-or-miss reliability and its shortcomings.
  • Note that if the wind info is of the utmost importance to you, the main choices are: Weather 14 Days, YR.no, FORECA and MSN Weather.
  • Because of dismissing of all the reasonable data sources, but also because of their own bugs, let’s specifically ditch the following apps:
    (1) Morecast has a rather peculiar design, and it only gets worse;
    (2) Weather Timeline can show outdated readings for all data sources, no matter how often you’re forcing an update (tested with Dark Sky, Weather Underground and YR.no as data sources); its widgets bore me to death; and the app itself is pissing me off;
    (3) Today Weather seems well-designed at first, but in the end its widgets don’t satisfy me, either because of their design, or because the icon shows rain for the entire day even for a meager 20% chance of light rain.
  • Bugs that lead to “hesitations” of ±1°C (jumps up and down; or lack of sync between the app, the notification, and the widgets; or differences between the automatically determined location and the manually entered one) affect the Android apps for: The Weather Channel (including when used in Google News & Weather, Weather Timeline, Today Weather), Dark Sky (with further delays when used by Forecaster, Weather Timeline, Today Weather), sometimes Weather & Radar Pro (I noticed this after updating from 4.8.1 to 4.9.2), and… AccuWeather (but its’s a piece of junk anyway).

Fuck me sideways if I know what app I should trust!

[39]IMPORTANT! No matter the app, when you’re using an apps’s widgets and you want them updated with the configured frequency, make sure the respective weather app is not battery-optimized in Android 6 or newer:

There are widgets that have a symbol for manually updating them on demand, but even such updates would fail if the corresponding app is “optimized”!

Note that configuring a weather app to display an ongoing notification in the system notification area might or might not help: I’ve seen cases when the notification was not updated in due course even after manually opening the app, while the widget was kept up-to-date!

[40]—Originally left as a placeholder for further additions, this number just wanted to be filled up. As I monitored the extremely volatile weather on the end-August in this part of Baden-Württemberg, I became more and more disappointed by all of the weather apps.

All the weather sources were agreeing that the original forecast (see DWD’s charts at [37]) was wrong, and that there will be absolutely no rain on Friday, Aug. 25. Well, guess what? Between about 4:45 and 5:01 AM there was a sudden shower predicted by nobody!

And yet, the radar maps in Weather & Radar Pro, DWD and wetter.com were showing it:

After the rain started, wetter.com somehow got a clue about it, but just as it was ending, it thought the rain would still last for 5 more minutes. It then switched back to a non-raining icon:

The new ability of wetter.com to announce the expected rains that are not part of any official warning is far from being accurate. There were other warnings, as the one from 17:00 the same day (“Little precipitation expected in 70 minutes”) that wasn’t really supported by the map, and the rain never happened. Also note that the rain that the rain info shown in the above pic (“Light precipitation ends in 5 minutes”) might be purely serendipitous: the temperature and the condition were “calculated at 4:45” and shown at 5:00, but it’s not uncommon to see values calculated 30…60 minutes ago!

Too many false rain alerts by all the weather apps for the respective weekend, including RainToday with messages like this one: “Rain will start in 11 minutes and will last for 7 minutes.” Well, 12 minutes later… “No rain expected.” And yet… some more 10 minutes later, a 2-3 minutes shower just took place!

Or, as you can see below, “Rain will start in 48 minutes,” but only 4 minutes later… “No rain expected” for the next hour! Kaboom.

Predicting rain is a nonsense, and whoever designed these apps was dumb. They don’t even say “A rain might start…” but “Rain will start…”–how the fuck can they be so sure when in most cases they fail?!

I should also say that during the respective I was particularly disappointed by: Weather 14 days, for it predicted too much rain that never happened, and because it showed temperatures that were almost always either too high or too low–the proof that it always displays forecasts and never measured valuesAccuWeather, which I wanted to take a look at both via its official app and via 3rd-party apps that were using it; Dark Sky, for it always displayed temperatures that were severely off the charts.

Needless to say, this never materialized:

Being disappointed by everyone doesn’t help. What next?

[41]—It’s the time to explain how I gave AccuWeather a chance to prove me wrong in my belief about them. I simply tried GO Weather Forecast & Widgets (aka GOWeatherEX) for a while, with the cute Peanuts Weather Widget Theme and Peanuts Weather LiveBackground.

The app doesn’t look that bad, but the default theme is yet another crappy showcase for random background pictures:

Let’s try Peanuts:

The background changes depending on the weather and the time of the day:

There are actually 14 possible backgrounds, 7 for daytime and 7 for the night: cloudy, foggy, overcast, rain, snow, sunny (clear, because “sunny night” is crazy), thunder; a couple of them are animated.

The widgets don’t change their backgrounds, but there are several types of widgets nonetheless:

Vicariously experiencing AccuWeather via GO Weather wasn’t such a bad experience, but the data was really bad. Maybe the current conditions weren’t that dumb as those displayed by Dark Sky, and they were sometimes even better that those shown by Weather 14 days, but still.

To make things worse, the readings were behind those displayed by m.accuweather.com. I already noted that AccuWeather‘s own app can lag behind the website, and forcing a refresh won’t help. The same is true for GO Weather, but this app can lag even behind the official app, which sometimes means a guaranteed delay of 30 minutes until this app would show what’s on the website. The same delay affects Today Weather when it’s using AccuWeather as a data source. Unacceptable to me.

I could come with a hypothesis though. I’ve read somewhere that The Weather Channel and Weather Underground are not supposed to give the same current conditions, but the forecasts should be identical, and when they’re not… “it’s because of the way they interpret a location”! Maybe AccuWeather and the apps using it can all display “Leonberg,” but depending on the app, data for a different exact location might be computed! (Technically, Eltingen and Höfingen are also Leonberg and they all share the same ZIP code, yet they can show as distinct locations in some databases.)

Other bugs in the GO Weather app:

  • One should preserve the default date format yyyy/MM/dd, otherwise the multiple-days widgets won’t display anything for the next days.
  • The impossibility to manually add Böblingen or Boeblingen. Unless you’re physically there, you must add Sindelfingen instead.
  • The impossibility to manually add the correct Leonberg. Searching for it finds Leonberg, Bavaria and Leonberg, Baden-Württemberg. Then, no matter which of them, it will add the Bavarian one!

All in all, I can’t recommend GO Weather to anyone.

…nor could I recommend the official AccuWeather app, being it the paid AccuWeather Platinum edition. Even the latest 4.9.1-paid edition has the most stupid widgets possible. Say you want to be warned when to take an umbrella, but also to have an overview of the next couple of days, so you’d be using this widget:

The widget shows the current temperature, but not the sky condition! Is it sunny, cloudy, does it rain? The only fix is to also use the ongoing notification:

If you only want to use a widget that also shows an icon for the current condition… or at least a text, no warnings to take an umbrella or a jacket!

If there’s any consolation, for the day shown in the above widgets there was no real need to take an umbrella–the forecast was a bogus one, as it’s often the case with “the most accurate weather company worldwide”…

BONUS: A rather nice OEM app that uses AccuWeather is Weather – Simplicity Weather, by mie-alcatel.support, which doesn’t need an Alcatel phone. Obviously, this is yet another Chinese app…

[42]—Speaking of bugs and out-of-sync situations, let’s revisit The Weather Channel and Weather Underground, both as official apps and via proxies.

(1°) Did you know that there are two official The Weather Channel apps, but most people can’t see one of them if they’re connected to their Google account? Here:

  • The “normal” one, at version 8.0.1 as I’m writing this (I couldn’t notice any visible changes over 7.14.2 or even 7.11.0), bloated with ads, is what most people would see. Android name: com.weather.Weather
  • The “hidden” one, at version 1.12.01 as I’m writing this, much more lighter (a 4 MB download!) and having a different design, is probably limited to some older Android versions. Android name: com.weather.alps

The smaller one isn’t listed in most places, e.g. it’s not listed on APK Mirror, but it can be downloaded from apkpure.com.

They’re not only different in look, they can be out of sync with each other! Manually forcing a refresh doesn’t help when “the small one” persists in not updating itself!

Note how “the small one” mentions that a rain is possible at 14:00, yet “the smart one” can’t be bothered with that.

Another possible rain mentioned by “the small one,” even as it displays outdated information: the rain shower (which never happened, mais passons) changed into “mostly cloudy” and a higher temperature in the full app.

Both apps are a PITA in different ways, so I can’t wholeheartedly use any of them.

(2°) Let’s see now a bug in Today Weather using weather.com as a source (yes, I refreshed both apps more than once):

Today Weather gives as chances of rain: 0% for Tue, 40% for Wed.

The Weather Channel gives as chances of rain: 10% for Tue, 10% for Wed.

Everything else is identical. What the fucking fuck?! (I’ll add that the two apps don’t agree on the usage of the icons that include rain, so reading the chances of rain is important.)

(3°) Weather Timeline using Weather Underground, because everyone loves both of them. At some point I tried to see the difference between Leonberg and Böblingen, towns at 13 km from each other as the crow flies and 20 km by road. Ouch… everything is 100% identical: the wind, the humidity, everything!

Trying to read the correct values in Weather Underground‘s official app I noticed that:

  • The GPS-provided location for Leonberg was Eltingen, which is still Leonberg, but 1-2 km to the South.
  • Manually selecting Leonberg displayed… Leonberg, Bavaria, not Leonberg, Baden-Württemberg!
  • Additionally, the second PWS was displaying garbage: the weather station was probably in plain sun.
  • Böblingen or Boeblingen don’t exist if one tries to add them manually! I had to add Sindelfingen, the sister town that’s 3 km away.

I told you that Weather Underground is a piece of crap.

(4°) Oh, but Today Weather too can use Weather Underground! OK, let’s see:

I’m not sure what is this app reading. At first sight, the same location, but no! The humidity is not the same (49% vs. 51%), and the forecasts are different (28-27-25-23-22-21°C vs. 27-25-23-22-21-21°C). So maybe this app somehow manages to get it right.

But how can I trust an app that has its own bugs with a data source that uses highly questionable PWS?

Back to square one.

[43]—A late discovery, Climendo Basic and Climendo (ad-free), also climendo.com, a strange forecasting apps that compares different providers. They list 9 providers, but unless you’re in the UK, in Norway, Sweden or the US, count on 5: AccuWeather, Dark Sky, FORECA, WorldWeatherOnline, Weather Underground.

The way they estimate the reliability of the prediction based on how much different sources agree, and the way they compute the estimation when the sources disagree severely, are a bit fuzzy: if we are to make an average of the 5 forecasts, in the first screen the temperature should have been 19°C instead of 18°C; and in the second screen the amount of rain should have been 0.6 mm, not 0.3 mm!

Other types of info:

Apparently, Weather Underground is the most reliable source for Leonberg. Things can be very different though:

Unfortunately, their approach of assessing the accuracy is wrong:

  1. If for a location Dark Sky or FORECA or WWO or any other source is clearly divergent from all the others for a given location and time, why should Climendo compute an estimation using this “rebellious” source whose only effect is to spoil the result? I’ve seen “fairly certain” estimations where one of the sources showed a very different temperature.
  2. Accuracy is only computed for temperature and wind. But one of the most important role of a weather forecast is to predict precipitation, and this aspect is ignored in the accuracy screens! Who cares if a source guessed the correct temperature or the speed of the wind if it missed a heavy rain or heavy snow? Moreover, unlike the temperature, the exact speed of the wind is irrelevant for most purposes (even for airports!); what matters is to know whether it’s gonna be 7 km/h or 80 km/h! (This is why they invented the Beaufort wind force scale.)
  3. Then, just as I pointed out at AccuWeather (see [16]), to determine the accuracy one needs actual measurements, but there aren’t official weather stations for all of the locations! In such cases, computing an accuracy based on unreliable PWS is absolutely pointless.

As a proof for the last point, observe in the last screens Ghimbav and Poiana Brașov. They’re both Brașov practically, but Ghimbav has an official weather station (too bad not all the weather sources use it!) that’s used for Brașov; whereas Poiana Brașov doesn’t have anything, and because of the different altitude, it usually features a different temperature. To say that Weather Underground is the best forecaster for the temperature in both locations is stupid once you know that it only uses PWS and no official station. To say that the wind was best predicted by AccuWeather for one location and by Dark Sky for the other is playing dice: they both use mathematical models applied on satellite maps, but whatever Dark Sky says cannot be compared to any reliable reading by a weather station! There is no reliable PWS in Poiana Brașov, full stop.

All the bloody statistics made by Climendo are crap. They collected some aggregated data and they just played with them, but the results are meaningless.

What’s worse, not only the app lacks a widget, but I can’t find it useful in estimating the weather for the next days. Somehow, it’s not practical.

[44]So what are the weather apps that I’m still using? I’m writing this, I keep installed:

  • Weather & Radar Pro Ad-Free (a 2.99 € value) → for the overall experience and the radar maps
  • Weather 14 Days Pro (a 1.99 € value) → especially for the forecasts, as a second opinion
  • DWD’s WarnWetter → warnings, official readings for the closest weather stations, short-term forecasts (only for Germany!)

But this is not a label of quality by any means. And yet…

Weather & Radar Pro can be used not only in Germany. Its Android version is available worldwide–unlike its Windows 10 and Windows Phone (WP 8.1) editions, which are only available in Germany (or maybe in DE-AT-CH aka DACH) and in German. Once we know that no weather app can be very reliable, we might accept it as the best compromise: the widgets are nice and useful, the app is quite informative, and it has a weather radar that works. Here’s some French reviews on Google Play (for the free edition Météo & Radar):

[45]—In a morning that proved to be chiller than predicted (13°C measured at Renningen-Ihingerhof, whereas the best prediction was 14°C and the forecast-only sources predicted: YR.no 18°C, Weather 14 Days 16°C), I decided to give another chance to the French private network METEO CONSULT and their app Weather Crave / La Chaîne Météo. They do offer forecasts for the entire world, and even for France, they only show forecasts (for 204 locations).

Notwithstanding its unsatisfactory widgets and its questionable usability (despite the apparently elegant design), it might be of some help to some people–if they really prefer forecasts for the current hour instead of possibly unreliable actual readings.

The “Live” section and widget are also forecast values. When it thought there were 23°C with “Chill factor 32°C” (wow, FAIL!), there were rather 21°C with “Feels like 23°C.”

What I learned is that this apps compares several forecast models: its own one (Weather Crave, by Meteo Consult), the French one (ARPEGE, by Météo France), and the American one (NOAA’s GFS).

In this particular day, none of them was particularly accurate.

If this, and the fact that one needs to scroll to much to get the relevant info, is not enough to deter you from using this app, take this one. While Meteo Consult is decent enough to show different forecast data for Leonberg, Renningen, Rutesheim, Gerlingen, Stuttgart, and the airport, they also have another site, La Chaîne Météo, and… the readings for Leonberg (the correct one, ID=71428, not the one from Bavaria!) are different!

Not by much, but still…

  • Maxima on MeteoConsult.fr: 28-31-23-18-17-20
  • Maxima on LaChaineMeteo.com: 27-30-24-18-17-20

As a general rule, I can’t trust them.

Since we’ve visited a French app, and mentioned the ARPEGE forecast model, let’s make clear that Météo-France is not very well suited for locations outside France, unless you’re in a big city. In my case, Stuttgart is the closest location it knows about. And the widgets are buggy. But to name the worst thing, I’ll quote Fritz Müller: “Déinstallé immédiatement. La publicité est insupportable.”

If you want to explore the ARPEGE model, Quentin Lagarde’s Météociel can show you maps from lots, really lots of models:

Unfortunately, this is the 3rd French app that has a design based on in-app tiles, and it sucks even more than the two apps above; and the widget is really bad.

[46]—And something curious I’ve just discovered: two apps by CH Mediendesign. The most usable one is simply called Wetter, and it uses AccuWeather for data. As expected, despite any manual update, it lags a bit behind m.accuweather.com, and the weather data itself is not in sync with the reality, but the app is very informative in a single screen:

The widget I prefer is as informative as the optional notification (which however also shows wind info):

UPDATE: There above app is not an original one, but a clone!

First, I found a clone of it called Weather Forecast Widget Clock (Animated Weather), by Apps for Droids (from New Zealand): 100% identical, except for an extra banner with ads at the bottom! Of course, it’s unclear who stole whose graphic resources. Then, I’ve found another clone, from Morocco: Worldwide Accurate Weather, which in the About box it links to the original source of all these clones, which is… the Chinese Geometric Weather, by WangDaYeeeeee!

Now it starts making sense that the guy behind CH Mediendesign prefers a 2nd app, Wettervorhersage, which has an ad-free Pro version and a dedicated website. This 2nd app uses data from Dark Sky, which sucks outside US/UK, and has two other major flaws: no widget, and a non-ergonomic home screen that need to be scrolled to see what is to be seen–which is much less useful and even more difficult to read than in the previous app:

Oh, one more thing: in their Wetter app, what’s this Polish Badenia-Wirtembergia thing doing in a German app that tries to be partially localized in English on English-language phones? We got Germany, but then…

[47]—The above pic reminded my that I forgot to address some unexplainable incongruities found in weather apps.

Both CH Mediendesign’s Wetter and the app Weather 14 Days consider the day as between 0:00 and 23:59, unlike The Weather Channel, for which a day starts at 6:00 AM and ends at 5:59 AM the next day. And yet, their widgets still include the icon with the maximum and minimum for the previous day after midnight (probably until 6:00). In the screens below, we’re at 1:49-1:50 on a Wednesday morning, but the first icon reads DI or TUE, i.e. Tuesday:

Well, what I just wrote above about the first app is actually wrong, and I only got it in the morning. Take a look at the following pics, where the forecast helps a lot, as today’s max/min of 34°/23°C is followed by colder days, with tomorrow’s min/max of 23°/16°C:

What happens here is that the minimum “for today” happens to be the maximum for tomorrow, and tomorrow’s minimum (16°C) is below the minimum “for today” (23°C), which actually takes place… tomorrow morning. AccuWeather doesn’t release hourly forecasts for more than 24 hours, but the trend is clear: 16°C will be met after-tomorrow morning, and so on. Each time, the minimum for a day is actually to be observed… the next morning or, in AccuWeather’s terms, “Overnight.” Basically, AccuWeather is as stupid as The Weather Channel: those Americans can’t fucking understand that a day ends at midnight!

In the first screen above you can see how m.accuweather.com splits the day in Morning, Afternoon, Evening and… Overnight (which is the next day). But then when The Weather Channel gives two temperatures (e.g. 18°C and 12°C), the first one is for Today, and the second one is for… Tonight (again, this is actually tomorrow). Morons. Early morning is too much for them. But wait! America is totally screwed: Merriam Webster’s Learner’s Dictionary gives this explanation for Parts of the Day:

Morning 5 am to 12 pm (noon)
Early morning 5 to 8 am
• Late morning 11 am to 12pm

Afternoon 12 pm to 5 pm
• Early afternoon 1 to 3pm
• Late afternoon 4 to 5pm

Evening 5 pm to 9 pm
• Early evening 5 to 7 pm
Night 9 pm to 4 am

The nice part is that between 4 and 5 AM there seems to be nothing, nada, zilch, void. (If you go up to [27], you’ll notice that for wetter.com, between 5 and 6 AM there’s nothing, no part of the day!)

Here’s how Morecast–an app made in Austria–properly divides the day in: Night (0-6), Morning (6-12), Afternoon (12-18), Evening (18-24):

As for Weather 14 Days, I’ve checked with the hourly forecasts, and the minimum temperature is indeed something that happens between midnight and 23:59 the same day; this doesn’t explain though why the widget kept using yesterday’s icon after midnight!

A different kind of incongruity related to hourly forecasts affects however Weather 14 Days (among other apps):

We’re between 1 AM and 2 AM, which have forecasted 18°C and 17°C for Leonberg, yet the current temperature is 20°C; for Böblingen, the hourly forecasts for 1 AM and 2 AM are 17°C, yet the app displays 18°C!

Unless I’m terribly wrong, this app cannot have the excuse of displaying measured current temperatures, because it simply doesn’t do that! The fact that it always displays forecasts for the hour makes it so ridiculous sometimes. To the point, the actual temperature was more like 19°C rather than 20°C, or even 16.5°C measured 8 km away (Ihingerhof). The occasional presence of a different current temperature than the forecasted one is a mystery in Weather 14 Days; it normally sticks to the forecast, no matter how wrong it is (à la YR.no).

Of course, with apps that do display measured current values (even readings from weather stations too far away, as FORECA and MSN Weather do), which basically means “all the other apps excepting those who use YR.no,” we can have obvious reproofs:

  • Why don’t they change the hourly forecasts, even only in an improvised way, if the current readings flagrantly contradict the forecast? I often see weather apps saying: “it’s getting warmer and warmer today, there are now 32°C already, and the forecast for the next hours are 29°C, then 30°C, then 31°C”!
  • Why don’t they change the daily maximum or minimum if the current reading topped the scale? So you can have currently 32°C, but the maximum for the day still reading 29°C.

A third kind of incongruity relates to the maximum for a day never being forecasted for any given hour! This affects many apps (and it’s making me crazy!), but let’s start with the one we’ve been just talking about:

The 31°C maximum is never to be met at any given time; between 1 PM and 4 PM, only 29°C are expected.

As the forecasted max/min for a day can change sometimes several times a day, it actually changed 12 hours later, to 30°C–but the hourly values didn’t budged.

But let’s inspect hourly forecasts further in future, maybe they’ll make more sense:

Nope. Tomorrow’s 22°C won’t be met at any point! WTF?!

Looking farther in the future reveals the volatility of the forecasts. Bucharest, Sept. 12 (five days in future): 40°C just out of the blue, although at around 3 PM the forecast is for 31°C! Some 16 hours later, the forecast for the day drops to… 25°C (that’s huge!), which might happen somewhere around 9 AM (23°C):

For all such inconsistencies–everything described at [47]I blame the software developers, not the meteorologists.

[48]—By popular demand (just kidding), I decided to reluctantly try YoWindow Weather (also ad-free), simply because I’ve heard of many people who love it.

The official description is wrong (probably outdated): “The weather forecast is provided by yr.no and NWS – the leading meteorological organizations.” In fact, there are several other weather data sources, and they can be found under Location settings, not under Settings:

The defaults for my location were: METAR (the airport) for current conditions, and met.no (YR.no) for the forecasts. Frankly, all the options for current weather are poor (FORECA would show data for Stuttgart / Schnarrenberg or for the airport); so I had to choose Weather Underground, which means PWS.

Trying to manually select the weather station reveals that they can only read the current conditions from (1°) airports, and (2°) various PWS (which means Weather Underground). Should you opt for e.g. for FORECA, there’s no way to select a specific weather station only known to FORECA and not to Weather Underground.

Almost 4 screens of (mainly) PWS–which one to prefer, really?!

The funny thing is that if you select anything but an airport, it warns that PWS are not reliable (Pulcinella’s secret):

But should you merely select Weather Underground for current weather, and leave “Weather station” to auto…

Of course it uses a PWS, what else could it be using? But there’s no warning that Weather Underground is intrinsically unreliable!

Otherwise, the most complex widget shows–with a choice of two icon sets–conditions for 5 days, with 3-hour interval details:

This is however incomplete information, because the icon for any day only shows the maximum temperature, without any minimum–I expect this to be an annoyance in the winter.

The same info, plus an animated (and customizable) background (hint: metric units with pressure in mmHg is called “Russian”):

Oh, wait! This app doesn’t show the maximum temperature for the day, but for the “American day”–that is, for the day after 6 AM! In the example below, the temperature of 18°C expected between midnight and 6 AM can’t help the maximum for the day to be shown as 12°C, because the temperature decreases during the day:

What a crap.

[49]—I mentioned a lot that my city, Leonberg, has a part called Eltingen; that Gerlingen is a nearby town; and that Deutscher Wetterdienst has an official (albeit mediocre, without wind and rain info) weather station in Renningen-Ihingerhof. Here’s a relevant map, where:

  • Leonberg is everything in the red border; the bottom half is Eltingen; the width shown on the “border” between Leonberg proper and Eltingen is 3 km
  • 1 is my place, in Leonberg Altstadt
  • 2 is what long time ago used to be Eltingen Altstadt, but now Eltingen is part of Leonberg
  • 3 is Gerlingen
  • 4 is the location of the weather station Renningen-Ihingerhof, actually between Renningen and Magstadt
  • 6 is Höfingen, part of Leonberg, but which could have been a distinct village

Despite the official distance between Leonberg and Renningen being 8 km “as the crow flies” and 12 km on road, one can see that everything seems to be closer. I can have a reasonable trust in the hourly measurements made by the station at 4, right? And to test various apps, I can set my place to Eltingen, Gerlingen. even Rutesheim or Renningen.

And this is where I show you how Weather Underground fails. I didn’t take screenshots of the official app, but here’s YoWindow using WeatherUnderground with Eltingen (left) vs. Leonberg (right):

I did this because the official Weather Underground app locates me in… Eltingen, and if I try to select Leonberg, it would show the wrong Leonberg (the one in Bavaria!)–see the screenshots at [42](3°). So it was better to compare what other apps that are using Weather Underground‘s API are getting–garbage! 3°C is too much of a difference. As a general rule, Eltingen seems to be “hotter” than Leonberg.

[50]—Mentioned at [18],  Transparent Clock & Weather can use: FORECA; met.no (which is YR.no); Weather Underground and OpenWeatherMap. The app seems to have gained some faithful users, so here’s how it looks like, with its 6 tabbed screens (click to enlarge):

Usability-wise, notice how the hourly forecast (screen 2) is made more legible thanks to the line chart (screen 3), whereas the daily forecast (screen 5) is only a dumb list. The wind direction is absolutely irrelevant if you’re not a pilot, and the moon icons that don’t show the illumination percentage for each day are pure idiocy.

The icons seen above are not the default ones; they have 11 sets (and one can add even more sets from here, here, and here), and I chose Vivo:

The set I opted for helps me point out an inconsistency in their widgets:

This is not the only app that has such incongruities, but usually it was the other way around, i.e. smaller widgets show the icon for the current conditions and forget to mention anything about the current day (min/max and icon); notice that the 4×2 widget has Friday (=today) as the first day in the bottom row, whereas the 4×1 widget starts with Saturday (=tomorrow). I’m pretty sure the developer wanted to start the smaller widget with the icon for “now” (because it’s larger and near the temperature), but got the things mixed up. note that towards the end of [41], I illustrated how AccuWeather’s largest widget (the one mentioning an umbrella) forgets to mention the current condition beside the temperature, while the other widget does it as a text (“Sunny”).

[51]—Here’s the latest app I discovered, “the exceptionally useful no-nonsense weather app for iOS and Android” called… Hello Weather! Fabulous graphics, obviously developed with iOS in mind (no widget, by the way); using Dark Sky by default, can also use Weather Underground (paying users).

It’s not just night theme vs. day theme; the colors also depend on the weather (temperature):

Note the advantages of being a paying user (unfortunately, $2.99 translates into €3.29, and $6.99 into €7.49):

It’s not impossible for this app to add some other weather data sources in the future–in the VIP edition, most likely.

UPDATE: Here’s the difference between the sources Dark Sky (left) and Weather Underground (right):

They’re both more or less wrong:

  • There was a strong wind, as forecasted by Dark Sky and by everyone else (there was an official warning of wind gusts)… but the PWS used by Weather Underground was just crapola for not measuring any wind speed!
  • The temperature was more like 18°C, or closer to 17°C anyway.
  • A sudden rain shower just ended, so there was sunny enough, but with heavy clouds racing at high speed. I’m not sure what the right icon should have been, but both sources show no chances of rain, which is bogus! (The Weather Channel also failed to predict any drop of rain.)

I’d say that the icons and the other conditions (temperature, wind) displayed by Weather & Radar Pro and Weather 14 Days were the best ones for the hour (and wetter.com changed its icon during the hour from “light rain” to “cloudy”–which is sunnier than “overcast”). For once, the weather was exactly as forecasted by several European sources, and Weather 14 Days expressed the forecast as follows:

Note the wind speed and the small amount of rain. Just as it happened.

Oh well, shortly before 15:00, Weather Underground acquiesced “27 km/h winds from the west” (and a raise to 18°C; whereas Dark Sky downgraded to 19°C)–almost 2 hours after anyone could experience the strong wind gusts! Oh my… better late than never, they say.

But then of course, around 23:45 there was a rain shower, and both sources used by Hello Weather fail to predict it, as the risk of rain was zero for the hour and for the next hours! Most other apps failed lamentably too. The Weather Channel was the only one to properly show a rain icon at about the proper time; wetter.com showed no rain in the icon, but the text “Rain will stop in 5 minutes” in the app, and after the rain stopped, the icon changed into a light rain one (since version 2.20, the “calculated” current conditions are disconnected from the new contextual information). Oh well.

One more discovery about Hello Weather: the lifetime VIP membership is $6.99 (or €7.49), but should you pay a 1-yr membership for $2.99 (or €3.29), the app will state that a lifetime membership only costs $2.99 (hence €3.29); that would save you $1.01 or €0.91, so don’t jump to purchase the lifetime directly!

[52]WRAPPING UP:

  1. Generally, The Weather Channel (weather.com) is a very decent weather source for most places, but the official app is too loaded with crap, videos and whatnot. The few unofficial apps that can use it (Google News & Weather, Weather TimelineToday WeatherWeather WizChronus) have each their own issues, which makes me reluctant to recommend any of them. Also note the “American thing” of considering that a day ends the next morning at about 6 AM (see [9] and [47]; the issue also affects AccuWeather).
  2. No, AccuWeather is not reliable enough in my opinion. Also, its own app lags behind the official site. And sometimes the site too seems to be serving old data for the current conditions. Same for the apps using it (e.g. GO Weather Forecast & Widgets, 360 Weather, Today Weather, and more).
  3. Weather Underground, despite being bought by the company behind The Weather Channel, has different readings for the current conditions (despite the forecasts being basically the same; sometimes the location is interpreted differently though). The problem is that they use Personal Weather Stations (e.g. cheap Netatmo boxes anyone can install under the roof) that have absolutely no dependability: several PWS situated two streets away from each other can report drastically different current conditions, and they typically report higher temperatures than normal, for not being placed in proper conditions (they get heated by the sun or by the building); the reported values might also be too old. The official app (which is not bug-free!) allows you to select a different PWS, but 3rd-party apps are served with the readings for the default PWS for the respective location. Apps using it include: Weather Timeline, the incredibly annoying yet original Morecast, Today Weather, Weather XL Pro, Chronus, MacroPinch’s Weather, YoWindow Weather, and the paying version of Hello Weather.
  4. Yahoo! Weather is neglected by most, but it’s almost decent. Beyond the official app, it’s notably used by Chronus.
  5. Dark Sky, formerly Forecast.io, might be decent enough for US and UK+Eire, but for the rest of the world it only offers forecasts and no actual readings whatsoever, and the forecasts are less than stellar–so it can be quite off quite often. The official app has horrendous widgets and it’s not officially available outside US/UK/IE. What other apps are using this weather source? Two nice apps, Forecaster and Hello Weather; the newer Chill Weather + Widgets and Mosum; the ugly Arcus WeatherAmber Weather (formerly EZ Weather) and its sister Amber Weather&Radar Free (formerly Amber Weather Elite); then Weather Timeline, Today Weather, and Chronus.
  6. The forecast-only service YR.no is less than satisfactory, and the official app is even worse. It’s also used by Weather Timeline, YoWindow Weather, Today WeatherChronus.
  7. Tiempo.com (and 5 other sites for 5 other languages) offers the app Weather 14 Days which is also strictly forecast-based, but very much nicer; unfortunately, it can be as wrong as YR.no when it’s wrong–sometimes even a bit worse. I have to admit I’m using Weather 14 Days for their 14-day detailed forecast.
  8. FORECA doesn’t offer a much better forecast, but it appears to be a much better service because for the current conditions it uses actual weather stations–too far from my place though, and the data can be old if the measuring station didn’t upload new data; to the official app one could add MSN Weather, Weather & Clock Widget for AndroidTransparent Clock & Weather, YoWindow Weather.
  9. A different forecast–an average of several weather models–is offered by Meteoblue.ch; their app is a bit dumb, and their forecast for the current hour is oftentimes queer.
  10. WetterOnline.de is a great service, and their app Weather & Radar / WetterOnline (I recommend the paying edition, Weather & Radar Pro Ad-Free / WetterOnline Pro, it’s worth 2.99 €) is one of the best possible choices–obviously, not only for Germany (the Windows 10 version is only available in German though, and only in Germany or in DACH). Also, possibly the best radar maps. Still, at times I wonder what are their sources for the current conditions, and how much is forecast in their readings. The medium-term forecasts can be exaggerated at times, but they eventually change it. The forecast for the next 7 days is detailed, whereas for the day 8-14 there’s only a chart with temperatures and precipitation.
  11. MeteoGroup is famous and they have their own network of PWS, yet their apps WeatherPro Free, WeatherPro and RainToday are a disappointment: the apps are too complex, the paying edition only shows the next 7 days and an extra subscription must be paid for more; and too often the current readings don’t change for 2-3 hours because the measuring station simply didn’t upload new data! The same data is used by RTL’s Wetter.de (only available in Germany), with the same delay in coming with new readings for the current conditions (being it the temperature or the precipitation!), but the 14-day forecast is free.
  12. wetter.com and their official app are known outside Germany for many years, and the short-term and long-term forecasts have constantly improved. Unfortunately, very often and even for locations called “Wetterstation” or “Klimastation” you’ll find “calculated” instead of “measured” temperatures and conditions, and for no matter which location the current conditions jump from being quite recent (15-30 minutes ago) to being unchanged for up to 2h30. And yet, their 15-day forecast is useful (marked “Outlook” and less detailed for the days 8 to 15).

I won’t mention again any other weather sources or apps that are: “minor”; so US-centric (or made for France, Italy, Switzerland, Australia, etc.) that for the rest of the world they only read METAR data (aviation weather) and vague forecasts; restricted to a single country; or displaying meteograms.

The bottom line? No bottom line whatsoever!

[53]—I just realized that all my attempts to document the various shortcomings of the different apps and weather data sources are futile, BECAUSE THEY ALL SUCK BEYOND REPAIR. After some more days of very unsettled weather, I simply gave up.

ALL the apps failed to predict anything sensible. When most apps said it won’t rain in the next couple of hours… it did rain. When they said it would definitely rain… it didn’t. And all this, all day long, for several days in a row!

Of course, unsettled weather means now it’s sunny, now it’s cloudy, overcast, or broken clouds, now it’s a quick shower, now it’s sunny again, now it rains again. But for fuck’s sake, consider that those were days with heavy wind (gusts up to 60…80 km/h), and the radar was something like this, yet the same app that displayed the radar claimed there was no serious chance of rain:

Or–remember the wind!–consider this radar:

Nay, it couldn’t possibly rain. Not now and not in the next 2 hours, no matter how fast the wind was moving those clouds! (Really?!)

For fuck’s sake, when the weather is so changeable, so volatile, so unsettled, there is always the possibility to use some icons like the ones below–taken from Weather & Radar Pro / WetterOnline:

So you can convey the information that, despite some hours of sun, there’s going to be some rain (with the chance in percentage and the amount in the number of drops under the cloud!), possibly even thunderstorms, also with wind of some intensity (notice the lines on the arrows, and the color of the wind cone)!

Or, for the next hour or hours, the forecast wouldn’t include the number of hours of sun, but on the other hand the sun can be replaced by the moon when appropriate:

But even the app that is able to show so much information in a single picture didn’t show such icons for the current conditions and for the hourly forecasts! No, sir! It preferred to alternate non-raining icons with raining icons every 1-3 hours, with the wrong forecast almost all the time! Morons.

I am in the fucking proximity of Stuttgart, not on planet Mars, yet not a single weather app could reliably tell me when I do need and when I don’t need an umbrella for almost an entire week! This is more than unacceptable, it’s preposterous!

AccuWeather and MeteoGroup‘s apps were unsurprisingly off by 2-3 hours, but I didn’t expect them to be in sync with the rain. Even as AccuWeather promised 8-10 hours of clear sky, it rained more than once, and it wasn’t clear by any stretch of the imagination. The other apps were equally ridiculous. Dark Sky‘s “hyperlocal” shit and Weather Underground’s unreliable PWS were a sad joke: always forecasting the opposite of what happened, and always reporting ridiculous current conditions. I thought I was going to get mad.

Forecast-only sources, such as YR.no (met.no) and Weather 14 Days were also expecting rain at least 6-8 hours too late or too early, and temperatures up to 5°C too low or too high.

I want to also stress on a more specific lack of reliability of Weather Underground’s PWS: they’re notoriously unreliable with regards to the wind information. The PWS at Gartenstadt ILEONBER72 (used by default for Eltingen/Leonberg) can report wind speeds that change every couple of seconds between 1 km/h and 12 km/h when the actual wind can go up to 30-40 km/h!

Here’s in Hello Weather the sources Dark Sky (left) vs. Weather Underground (right); I was lucky to have WU report 9 km/h instead of the more usual 1 km/h, but note Weather Underground how cannot be trusted to report the wind as an important information to show in the main screen (Dark Sky does that!), despite the warning of “disruption” read from a DWD warning (Dark Sky is much more informative in such situations):

This doesn’t mean Dark Sky is accurate for Continental Europe–it’s not. Whereas the icon for the entire day tends to show rain… for most days (yes, for too many days in any given month or year!), the icon for the hour can show 100% sun when it’s rather 100% rain! Another thing that bothers me is the inadequate wording: 32-66 km/h is anything but BREEZY, and heavy rain is anything but DRIZZLE (I’ve seen “breezy” and “drizzle” so often that it almost gave me nausea).

Unfortunately, most weather apps don’t care at all about the wind, despite it wind being extremely important–especially as there are so many locations where strong winds are quite common.

Even Weather & Radar Pro / WetterOnline only shows the wind info in the forecasts, but never in the icon with the current conditions! Why is that so?! Even if the current speed of the wind be computed from maps or forecasted, and it still should’ve been shown!

There is one app that apparently cares about wind, but it’s not consistent about it: Weather 14 Days. Yes, it shows the forecasted wind speed for each and every day or hour, and yet…

Let’s examine some inconsistencies:

The most detailed widget does show the wind (W 48 km/h), but it doesn’t use any symbol next to the big weather icon. Also, the wind for the next days is only shown in a different 4×2 widget, an ugly widget that doesn’t use an icon for the current conditions and that shows 5 instead of 6 days:

Back to the previous “huge” widget, it also shows the amount of rain beside the umbrella (0.2 mm), but no chance of rain (this app doesn’t care about such a thing). The app does show the humidity though (look at the two yellow screens)… but there’s no wind info there! The wind info is given in the forecast of the day (the list below), and one must expand the details for the day to read the expected wind speed for the hour! (Or, of course, there’s also the widget that doesn’t show the umbrella and the amount of rain… and it also lacks the “feels like” temperature.)

Somewhat similar to Dark SkyWeather 14 Days also believes it should show some rain drops for almost all of the days, and although it also shows the sun and it has two cloud colors to distinguish between the degrees of cloudiness, it still disappoints. It disappoints further by only including the textual information for the coming hours (“Cloudy intervals for the next hours” / “Gusts up to 73 km/h for the next few hours”) in the app: what’s the fucking use of a huge widget that can show data for 6 days if it can’t warn me about wind gusts of up to 73 km/h?!

I can’t stress enough: all the makers of weather apps are too stupid to deserve any kind words and any consideration. All the weather apps are pure garbage. FULL STOP.

Shopping list:

  1. Weather frog.
  2. Swallows (“When the swallows fly high, the weather will be dry.”)
  3. Pigs (“Pigs gather leaves and straw before a storm.”)

OK, I need a farm.

OMFG, I just learned about Google’s weather frog! There’s even a Frog Weather Shortcut app. And here’s a funny bug:

In the middle screen (“Tomorrow”), the icon reads 40%, and the bottom text says 60% chance of rain. The right screen clarifies it: 60% is for today, 40% is for tomorrow. Being an American app, some parts of it consider today as “by 6 AM tomorrow morning,” whereas some other parts of it correctly considered today as being the day that just started at midnight.

One could actually notice in the detailed view how one day ends at 5 AM and the next day starts at 7 AM; as the details are given for 2-hr chunks (i.e. after 7 AM there’s 9 AM but no 8 AM), we can safely assume the boundary between “days” is 6 AM:

I’ll use the help of the Chambers Thesaurus, 4th Ed., to tell you what they all are: morons, fools, blockheads, fat-heads, dolts, dunces, dimwits, simpletons, halfwits, idiots, cretins, imbeciles, ignoramuses, dupes, stooges, butts, laughing-stocks, clowns, comics, buffoons, jesters; nincompoops, asses, chumps, ninnies, neddies, clots, dopes, twits, nitwits, nits, suckers, mugs, twerps, birdbrains, silly-billies, berks, charlies; wallies, jerks, dumbos, muppets, pillocks, prats, dorks, geeks, plonkers, gits, nerds, dweebs, nerks, cloth heads, dipsticks, goofs, kooks, tosspots; dickheads, knobheads, pricks.

[54]—By reading various reviews on Google Play, I realized that many people fail to understand that some weather apps exclusively offer forecasts, and no measured conditions at all! (Gee, the world is really full of morons.) They fail to understand that YR.no (met.no) only offers forecasts for any location outside Norway, and this is the reason it can be terribly wrong at times! The same stands for Weather 14 Days, no matter it’s Spanish and not Scandinavian; for the Swiss meteoblue; and for a few others.

The Scandinavians have more of such apps. The Finns have FMI Weather, which is quite decent, actually. If you scroll down in the app, it’ll also show a list of weather stations (under “Weather observations”), to let you have a real input to compare to the forecasts. An example:

  • Forecast for Leonberg for 01:00: 11°C, clear sky, feels like 9°C.
  • Weather station Stuttgart (Schnarrenberg), distance 13.9 km: 13.9°C at 00:00, cloudy (8/8).
  • Weather station Stuttgart-Echterdingen (the airport), distance 20.1 km: 11.5°C at 00:00, almost cloudy (6/8).
  • Weather station Mühlacker, distance 21.3 km: 10.0°C at 00:00, cloudy (8/8).
  • Weather station Kaiserbach-Cronhuet, distance 50.8 km: 7.5°C at 00:00, clear sky (0/8).

At the same time, YR.no and Weather 14 Days forecasted 10°C, although they’re not always of the same opinion (here they disagreed on the cloudiness).

There’s a queer thing with the app and the website of the Finnish Meteorological Institute. Their website and app cannot find places like Leonberg, Böblingen, etc. (so I had to search for Stuttgart if I am to enter the location manually), but when I am physically in Leonberg, the app suddenly “knows it” thanks to the GPS, and it shows a slightly different forecast than the forecast for Stuttgart. I don’t understand what exactly is it doing here…

In contrast, the Norwegian Meteorological Institute allows the entering (both on the YR.no website and in the app) of no matter what location (including Leonberg, Böblingen, etc.). Another limitation: the Finnish forecast for international locations seems to be updated only once a day, whereas the Norwegian one is updated twice a day. As for which one is more accurate… it’s impossible for me to say. Sometimes it looks like the Finns are better at guessing it–sorry, I meant forecasting the weather.

As a side note, the FMI Weather app cannot use two or more widgets at the same time (say, one with the following hours and one with the following days, or any combination of widgets): one of the widgets won’t be updated properly or at all. It’s definitely a programming bug. On the other hand, scrolling in the app is slow and hectic–especially the vertical scrolling. Too bad, as the app is otherwise not that bad.

But there’s more. The app Supersää, unfortunately only in Suomi, offers a side-by-side comparison of the forecasts offered by the Finnish Meteorological Institute (Ilmatieteen laitos) and FORECA. While the simple widget might give the impression that the two sources agree…

…they actually quite disagree, see the middle pane:

An important note: here, FORECA is only used as a forecast source! In contrast, ForecaWeather and MSN Weather use FORECA forecasts plus readings from the nearest weather station! As I noted at [34], for Leonberg they’d use the stations Stuttgart / Schnarrenberg or Stuttgart-Echterdingen, which are 14 km and 20 km away. So the “current conditions” are strictly forecasted values only in Supersää.

Supersää too has a problem in finding locations such as Leonberg, Böblingen and the like. One must physically be located in Leonberg for this app to be able to find about such a town!

Here too the forecasts are particularized for the automatically-detected location, but this app has sometimes a bug when using two locations: it doesn’t update them consistently. In the widgets shown below,

  • at 16:26, Stuttgart is shown for 16:00, Leonberg (GPS) is shown for 17:00
  • at 17:07, Stuttgart is shown for 17:00, Leonberg (GPS) is shown for 17:00
  • at 17:23, Stuttgart is shown for 18:00, Leonberg (GPS) is shown for 17:00

Note that no amount of manually opening the app and forcing a refresh helped. The app itself refused to properly update the displayed data for one of the locations as long as two widgets were in use!

There’s one more thing to be learned from the above sources: most Scandinavian sources (the meteorological institutes of Norway and Finland) and the Spanish Weather 14 Days don’t offer the probability of rain for international locations, but only the amount of rain. The only Scandinavian source to offer the probability of rain for all the known locations is the Finnish private organization FORECA!

A last note about the differences between the actual conditions in Leonberg versus Stuttgart or Stuttgart-Echterdingen (the airport). DWD provides distinct forecasts for each of these locations; for Leonberg, it’s actually the agricultural weather station at Renningen-Ihingerhof that’s considered (see e.g. [31]):

Yes, they can be quite different, as you can see. Sometimes the weather in Stuttgart can be quite dissimilar, despite the modest distance from Leonberg (more than 14 km, because the distance to Renningen-Ihingerhof should be considered; let’s say 20 km).

[55]–At [33], while I was mentioning a number of weather forecast models, this one popped up as quite original: “HIRLAM is a regional model used by many weather services. WZ offers the Dutch (KNMI) and Finnish (FMI) version. It covers Europe with 10 km resolution and is calculated 4 times per day with 48 h forecasts.” As I had the opportunity to test the Finnish model (in the apps FMI Weather and Supersää), I looked for an app that would use forecasts from the Royal Netherlands Meteorological Institute (KNMI).

Enter the apps by Surfcheck, aimed primarily at the Netherlands and Germany, but offering forecasts and current conditions for the entire globe. The apps of interest are:

  • dé Weerapp–only in Dutch, with a modern design, but only a radar widget for the Netherlands.
  • Das Wetter Live–only in German, with an antiquated design and difficulties in using the GPS; not recommended; the widgets require a 2.49 €/yr subscription.
  • Das Wetter in Deutschland–only in German, the modernized edition of the previous app; the same 2.49 €/yr subscription is required to use any of the widgets except for the DWD radar for Germany.

The most elegant app is still the first one (in Dutch); the last one has however not only some real widgets, but beside the official DWD maps, it can show a plethora of other maps from various sources, e.g.:

The major fault I find with these apps is that the current conditions are identical for a 50-60 km radius around Stuttgart, simply because for any location in the area any of the above apps would read the conditions from the Leinfelden-Echterdingen airport! To add insult to the injury, they’re unable to read the visibility (“Sicht”) from an airport’s METAR data!

Another set of failed weather apps…

[56]–While at [18] I simply stated that OpenWeatherMap and WorldWeatherOnline are crappy as hell, and at [20] I added a few words on WorldWeatherOnline, I didn’t add anything on OpenWeatherMap, and certainly I didn’t mention any app that only uses it as a data source (apps that use it as one of the data sources include Weather Timeline, Transparent Clock & Weather, Chronus, and the apps at [22]). Here’s a shortlist of OpenWeatherMap-only weather apps that are recent or recently updated:

  • Weather Now, by Charles-Eugene LOUBAO. Nice app with nice, elegant widgets, but with a strange thing in all four widgets: the two temperatures in small typeface at the right of the current temperature are not the max./min. for the day (I couldn’t figure out what they are). The cutest and neatest app of all five, despite the impressive number of bugs I found and reported. The weather icons seem to be those used by Google’s “weather frog” (see the end of [53]).
  • Weather forecast + widget, by AVA Development. Unfortunately, same icon as Stefan Jösch’s Chill Wetter + Widgets (they both used a 3rd-party icon).The app is quite usable, although it’s not practical: each set of data for a 3-hour interval takes 25% of the screen. The two widgets are also usable, but could have been more compact.
  • Lazure Weather, by Karebubl–sophisticated design, but it always gives the wrong values for all the locations–I tested by comparing the readings given by the app against the values on OpenWeatherMap.com, and they were totally different in temperature, cloudiness, humidity, pressure. I suppose the reason is the same as for the app at [11]. Instead of trying to copy other apps’ design (partially from Today Weather, which is not original either), they should have made sure they can display real, not bogus values! The effort spent on the elegant UI was a total waste.
  • I should have listed Buneme Kyakilika’s Forecast at [18], because it supports multiple data sources, but only OpenWeatherMap is free; Dark Sky and Weather Underground require an IAP. Each of the two 4×2 widgets should have been 4×1 (they’re poorly designed anyway, making them useless). The radar is useless, showing me Africa and Europe (it can be zoomed in, but there’s no location marked on the map). Among the bugs (which I didn’t report this time): knowing that OpenWeatherMap shows the same data for Leonberg and for Böblingen, it’s strange how this app sometimes shows different values, e.g. 16°C for Leonberg and 15°C for Böblingen.
  • Simple weather & clock widget (No ads!)–not that bad, except for the idiotic font that also spoils the best of their widgets! Another drawback is that this is a Web app that doesn’t query OpenWeatherMap directly, but a cloud server belonging to the developer, and sometimes the forecast for the next days might be outdated, as well as the current conditions (Info would report “Data update times” e.g. “App: 2017.09.28 22:28:00 / Backend: 2017.09.28 22:28:00 / Meteo: 2017.09.28 21:50:00” or in the main screen “Updated: 22:28 / 21:50”). Practically, the weather data is updated at their end only once an hour, at “ten minutes to…”.

I left “the app with the ugly font” last because I discovered a way it differs from the others. As OpenWeatherMap does not differentiate at the icon level between light, moderate and heavy rain, the icons for several hours in a row can show rather “heavy rain,” despite the rain being very light and the quantities minuscule (even less than 0.01 mm), yet the icon for the day might only show cloudiness! This kind of makes sense, but the maker of the aforementioned app judged “at the icon level” and thought of changing the icon for the day to show rain as long as there’s rain at any moment during the day!

Let’s compare the official OWM site to our app:

While it can be argued that the decision was a wise one and that the icon for the day explicitly reads “Light rain,” the problem is that the icon itself still suggests heavy rain! I suppose nothing would ever change in such apps unless OpenWeatherMap themselves start using differentiating icons.

Should they ever do such a thing, there’s one more thing they could do: differentiate between “mostly sunny,” “partly cloudy” and “mostly cloudy,” and they could find inspiration in Google’s weather frog:

OK, let’s drop OWM. This weather source is only a moderate match for my location anyway, but I wanted to explore it a bit. Oh, I forgot: one would expect Weather Timeline to be the most reliable app to use with OpenWeatherMap, as it’s a very popular app, so it should have most of the visible bugs smashed, right? Wrong: Sam Ruston couldn’t care less that OpenWeatherMap and YR.no don’t forecast the chances of rain, but only the amount, and he simply interprets “unknown chance of rain” as “no rain,” despite the forecast for the hour being rain!

So it’s both rain and no rain, at the same time!

Also, the ever famous Chronus, in version 8.5.2, simply fails to work with OpenWeatherMap. I added my own API key, it said it verified it successfully (yay!), but the widget read either “OpenWeatherMap can’t resolve the location,” (despite having selected the first of the 3 Leonberg locations) or “Can’t reach OpenWeatherMap at the moment” for a different location (the unambiguous Böblingen)–despite API calls working just fine.

I’m afraid the world is full of idiots, and most of them are software developers…

[57]–I bet most people don’t know how different can be the weather units customarily used in different places of the world. It’s not just “imperial vs. metric”:

Weather Units Temperature Visibility Wind Precipitation Pressure Pressure given for
1. Imperial/US (classic) Fahrenheit miles mph inches inHg Sea level
2. Imperial/US (modern) Fahrenheit miles mph inches mb Sea level
3. Hybrid/UK Celsius miles mph mm hPa Sea level
4. Metric/Canada Celsius km km/h mm kPa Sea level
5. Metric/Europe Celsius km km/h* mm hPa Sea level
6. Metric/Scandinavia Celsius km m/s mm hPa Sea level
7. Metric/Russia Celsius km m/s mm mmHg Location level
8. Metric/Romania Celsius km km/h* l/m² mmHg** Location level
*speed communicated in km/h to the public, but measured in m/s at the weather stations
**pressure communicated in mmHg to the public, but measured in mb at the weather stations

Strange thing, a relatively small country like Romania managed to mix the European values (sic!) with the tradition of some units from Russia!

[58]–I keep discovering weather apps; useless weather apps, that is. Launched in 2013 as a weather service, not as an app, Poncho got in the meantime an Android app and an iOS one, and expanded its area of coverage from NYC to the entire planet (except for some services that are still NYC-only). Unfortunately, the forecasts and the current conditions are very often wrong for Germany (initially based on Weather Underground, then on Dark Sky, they’re now keeping secret their weather data sources, but it looks like they’re still using Dark Sky), the widgets are dumb (why should I need two widgets to know both the current conditions and the forecast?), and their humor is cheap, repetitive and just pathetic. Yet another failure, too bad for the nice graphics.

[59]–Random late discoveries:

  • Minimal Weather, an open-source app using forecast-only data from the Finnish Meteorological Institute (FMI). No widgets.
  • MyWeather, by Clover Team. A Chinese app using Weather Underground. I cannot select Leonberg, not even Eltingen (which is known to Weather Underground), only Böblingen or Stuttgart are available (I suppose many other locations around the globe are unavailable). A couple of widgets seem broken, most of them work just fine… just not for Leonberg.
  • RxWeather is a sort of an open-source limited clone of Weather Timeline, using Dark Sky and nothing more. No widgets.

  • Wild Weather (1.89 €), is an older app I’d met once; it’s using Dark Sky, it’s not updated anymore and it has no widgets.
  • Of course, one could use Kustom Industries‘ KWGT or KLWP and add extra stuff to build a weather widget, but there’s too much trouble, especially on a smartphone (I suppose editing is more comfortable on a tablet). No way.

[60]—Meaningful updates in a couple of weather apps!

First, DarkSky’s official app (which I have to retrieve from unofficial sources, as my Google account doesn’t locate me in US or UK) got a new timeline widget in version 1.8! (The same widget design can be used as a notification, which is also nice.)

What’s great about this widget is that, unlike Forecaster‘s similar one, it really shows the next 24 hours, not “today” or “tomorrow.” With Forecaster, at 10 PM one would only see the forecast for the next 2 hours (despite the next day’s one being more interesting), but using the widget for “tomorrow” would make it pointless for most of any day (except for evenings). What’s less than great is the dull gray used to mark “clear”–here, the golden yellow used by Forecaster is by far the best possible choice. On the other hand, Forecaster’s widget doesn’t show the temperatures under the timeline (they use a separate widget for that), so overall Dark Sky‘s own widget is a much better choice–that is, if you like the shades of gray and if you can download the app.

Well, this being said, this data source still isn’t very reliable outside US, UK and Eire: it failed to forecast rain more than once in the last week.

The second big update of the season is the release of the redesigned YR app, which was for some time in beta. Version 5.2.15 of October 19, 2017, comes with the new design I noted on Sept. 15 (see at the end of [19]), plus new widgets: 4×2 with or without the hour, and 4×1. I obviously prefer the new large widget without the clock:

This doesn’t mean the new app is much improved: the widgets still don’t show the wind speed, and the explanatory text should have been more to the left to fit (they simply replaced the hour with this text, that’s why). But it’s a significant update nonetheless.

[61]–I suppose meteoblue is to have widgets soon, before Christmas. While the official version is still “Cirrus uncinus 78” (July 5, 2017), beta testers can download “Cirrus uncinus 82” (October 26, 2017), which has widgets!

The fact that they exist in this beta build doesn’t mean they can be used. Using a widget requires a subscription that would also remove the ads, but the beta version doesn’t allow any IAP!

Their recipe for success is pure genius:

  1. Add widgets in the beta.
  2. Lock the widgets behind a subscription.
  3. Make the subscription impossible (most likely because it’s a beta).
  4. Ask the users to test the impossible-to-get widget and get feedback.

I suppose the workaround would be to leave the beta, install the regular app, purchase a subscription, than join the beta again and update the app. Too much fuss: what if the widgets aren’t working properly? In the meantime, the resources used for previewing the widgets show 3 widgets on black or white background:

Obviously, the amount of rain in the first widget must be 0-2 mm, not 0-2 m. Où est la précision suisse d’antan?