Reviving an old thread: in the last few days (sorry I cannot be more specific), the place names have changed to something more sensible: either the name of the city, or of the suburb in which I actually am. Much better. So if the dev changed something.... well, it was the right thing to change!! 😁

@idefix37 said in Model vs. forecast:

They explain what they do, but the other ones don’t give so much details ... but all of them show a lot of advertising.

As you say. And these are the two main reasons I'm using Windy (well, two of the three actually, the third being a great graphical display): (a) Windy tells me what it is doing and where the data comes from (being a scientist, I tend to be rather particular about knowing how the data is processed !) and (b) advertising is minimal in Windy (even Meteo-France is worth, shame on them....).

@idefix37 said in Strange place names in Android widget:

GPS option is useful if you are travelling, but I prefer to use one of my Favorites as starting page (then it save the battery if you don’t use the GPS of your mobile)

Well, yes, that's probably what I'll end up doing while in town. But I do travel a fair bit, so I'll end up using the GPS. Oh, well, it does not matter so much if the name is a bit odd, as long as the forecast is good :-)
Yet as you say, this suggests that the name database is a bit odd... where is it coming from, OSM ?

Further investigation reveals that all the names I have seen correspond to places that have place = isolated_dwelling tag in OSM, and of course a name=XXX tag.
Something like place=city|town|village would be more useful, I think... ;-)

Another oddity is that in at least one case, the place name displayed is not the closest place=isolated_dwelling. There are closer places having this tag. So, there must me something I have missed... It may have to do with the details of how Windy connects to OSM (directly to the OSM database, or through some sort of intermediate export...)

You're welcome, keep up the good job :-)

Reviving an old thread: in the last few days (sorry I cannot be more specific), the place names have changed to something more sensible: either the name of the city, or of the suburb in which I actually am. Much better. So if the dev changed something.... well, it was the right thing to change!! 😁

And a last one - travelling on Southern France yesterday, the place names displayed were much more reasonable (villages, I think). It would appear that the issue resides with the database used (OSM extract...), that is somehow poorly tagged specifically in certain parts of the country...

I think I can confirm that the place name displayed is the closest place = isolated_dwelling. In the past day or two, the place name displayed has fluctuated beween 2 or three names, each corresponding to one isolated dwelling, and each roughly equidistant from where I'm sitting.

However, having played a bit with QGIS and Voronoi polygons, it appears to be a bit more complex. Apparently, the name displayed is never exactly the name of the closest place. Looks like either my position is incorrect (by > 100 m, which i think is unlikely), or the position of the places is wrong (could happen with projection errors?). Or some of the places are not considered, for some reason (the database used is a weird subset of OSM - we know it is weird anyway as it is ignoring villages and towns in favour of isolated dwellings...).

You're welcome, keep up the good job :-)

@idefix37 said in Strange place names in Android widget:

This should interest Windy developers

Well that's why I report it. Should I post it somewhere else ?

Further investigation reveals that all the names I have seen correspond to places that have place = isolated_dwelling tag in OSM, and of course a name=XXX tag.
Something like place=city|town|village would be more useful, I think... ;-)

Another oddity is that in at least one case, the place name displayed is not the closest place=isolated_dwelling. There are closer places having this tag. So, there must me something I have missed... It may have to do with the details of how Windy connects to OSM (directly to the OSM database, or through some sort of intermediate export...)

@idefix37 said in Strange place names in Android widget:

Yes, Windy uses Open Street Maps as database:
https://community.windy.com/topic/4649/how-to-correct-errors-in-the-base-maps

Well... OSM normally has sensible place names... although I suppose it depends on which OSM "layer" (or tag) is actually used, of course.

@idefix37 said in Model vs. forecast:

They explain what they do, but the other ones don’t give so much details ... but all of them show a lot of advertising.

As you say. And these are the two main reasons I'm using Windy (well, two of the three actually, the third being a great graphical display): (a) Windy tells me what it is doing and where the data comes from (being a scientist, I tend to be rather particular about knowing how the data is processed !) and (b) advertising is minimal in Windy (even Meteo-France is worth, shame on them....).

@idefix37 said in Strange place names in Android widget:

GPS option is useful if you are travelling, but I prefer to use one of my Favorites as starting page (then it save the battery if you don’t use the GPS of your mobile)

Well, yes, that's probably what I'll end up doing while in town. But I do travel a fair bit, so I'll end up using the GPS. Oh, well, it does not matter so much if the name is a bit odd, as long as the forecast is good :-)
Yet as you say, this suggests that the name database is a bit odd... where is it coming from, OSM ?

@idefix37 said in Model vs. forecast:

But for all weather app you find on internet be sure that the most part (almost 100%) of them do not involve human interpretation.

Thanks for the comment. I'm (mildly) surprised by this statement: if most of the weather apps just feed us the GFS data, how come they differ in what they show ? I have not run systematic tests, but played with 4-5 apps lately and I'm under the impression that they do not show the same forecast. Surely then the apps must do something more ? Interpolating perhaps, with different resolutions ?

(does windy interpolate between grid nodes by the way, or do we just get the straight model output ?)

Also, is there a way to find out what each app is doing, which model it is based on, etc. ? (short answer : probably not, unless they explain on their website and one takes the time to read it all, I presume ?)