Hi! I noticed someone, 9 years ago, uloaded a bulk of forest roads in one area from coarse data. The intent was noble, but the data uploaded was not classified correctly, not had surface information etc.
I took some time to add surface information with JOSM but I was wondering if this was the right thing to do.
Here is my changelist:
My dilemma is as follow:
i could improve the classification of those roads, but they are incorrect to begin with as they don’t match well the satellite imagery.
Sounds like a situation similar the to TIGER imported roads in the desert southwest of the US.
My approach was to take the easiest course which meant it was on a case by case basis. If it was easiest to edit the existing imported track to bring it into alignment with imagery that is what I did. But if it was easier to delete the existing and create a new then I did that.
In the case of these desert roads, the TIGER import had them tagged as residential but if they existed at all they were no more that tracks so any change I made had to be an improvement.
Which brings up my general philosophy on making edits in OSM: Each edit should be an improvement on what is there but don’t get hung up on making it perfect.
As a person who drives FSRs on a regular basis, do not trust satellite imagery, at least in the area I live. The imagery is often old enough that the roads shown have been overgrown to the point where even using the satellite image you can not find enough of the former road to even be a footpath. Other roads have been decommisioned with heavy equipment and huge stones and stumps now occupy the space which shows as a road in the imagery. My pet peave is that people keep adding non-existent roads that they see in the imagery. In some cases re-adding roads I have removed due to their non-existence at the present time.
Yes. In this case the shape is sometimes a coarse approximation. Here is an example where i overlay one of the said road segment on recent aerial imagery provided by the gov of qc (not bing or esri satelitte)
Lors de l’édition dans des milieux forestiers, la prudence s’impose pour bien informer les utilisateurs des données. Si on n’a pas eu l’opportunité de circuler sur ces routes, il faut se contenter de spécifier les routes bien identifiées sur la carte de base du gouvernement du Québec ( tms:https://geoegl.msp.gouv.qc.ca/carto/tms/1.0.0/carte_gouv_qc_public@EPSG_3857/{zoom}/{x}/{-y}.png ).
La classification des routes est importante pour bien éclairer les utilisateurs. highway=tertiary ou secondary représente des routes plus importantes dans le réseau routier. Dans le contexte forestier où des camions remorques sur-dimensionnés ont la priorité, nous ajoutons généralement des informations telles que limite de vitesse et priorité (exemple Way: Chemin du Barrage (970832805) | OpenStreetMap). Il y a ensuite les routes mineures (highway=unclassified) et résidentielles (highway=residential).
Cette édition comprend des chemins avec des caractéristiques très différentes. Prenons d’abord la route de Manawan (highway=tertiary) qui est une route principale. En parallèle, je vois Chemin du Lac-Charland (102639229) qui a été classifié highway=track. On observe de nombreuses résidences ou chalets le long de ce chemin. Selon la carte de base du Québec, ce chemin existe. Si les résidences n’ont pas accès directement au chemin Manawan, la classification du Chemin du Lac-Charland correspond donc à highway=residential. Si c’est plutôt un chemin en désuétude, on peut conserver highway=track. Si la surface est en mauvais état, on peut ajouter des clés pour décrire cet état. Pour le Chemin Marcil (102633871) de l’autre côté du lac, on peut indiquer highway=unclassified ou residential.
J’obseve de nombreux chemins qui semblent temporaires et créés pour la coupe de bois. Dans ce cas, la classification highway=track s’applique bien. Mais on évite d’ajouter ces routes à partir de l’imagerie satellite si on ne connait pas leur état sur le terrain. Ces chemins ne sont généralement pas entretenus après la coupe de bois et après quelques années, les ponts et ponceaux s’affaisent. Si on conserve ces chemins, il est sans doute prudent d’ajouter des caractéristiques telles que surface et smoothness.
Adding more details is usually a good idea, though adding surface=* only from aerials can get quite difficult and often is impossible as light colored asphalt and compacted unpaved surface can look very similar.
Yet, again, choosing the correct classification from aerials is tricky but the chosen highway=road if definitely better than highway=track if there is any doubt.
First, I would try to improve or reuse them. The Improve Way Accuracy and the “Replace Geometry” action from Utilsplugin2 plugin can be handy. Deleting should be the last option and if there are still some traces on imagery it is usually better to use a life-cycle prefix, e.g. abandoned:highway=*, instead.
If they’re still visible in imagery, mark them as abandoned:highway, rather than completely deleting them, as they will still show on the map to stop this from happening
No, besides the name of the value highway=unclassified is a classification while highway=road is the proper tag for a way/road with uncertain classification and could be all from a motorway down to a path.
On classifie une route selon sa fonction. L’image ci-dessus ne nous indique que l’état de la chaussée et son étroitesse et ne nous permet pas de voir sa fonction. Ce sont les attributs tels que tracktype, surface et smoothness qui nous permettent de décrire la chaussée.
Yeah I agree. This is why I am tempted to delete them. They are clearly on the database from an old bulk upload and are misleading given the points you mention.
In this image, the thin brown lines are from the AQreseau+ dataset that I overlay locally for comparison and the dark straight lines are from OSM, bulk uploaded 9 years ago:
