I hope that the beginning of 2023 is going well for everyone.
I’d like to introduce myself - my name is David, and I have picked-up the role of TomTom Community Manager for the United States and Canada from Courtney.
I’m reaching out to inform you that we have created a new MapRoulette challenge in order to add Surface to ways in California. To detect the issues in this challenge, we used Overpass Turbo to select Motorways where the Surface tag was not present.
In the description of the challenge you can find some additional information that will make it easier for you to solve the tasks.
Would you like to see this extended to other states?
Do not hesitate to contact me to talk about possible collaborations, problems or challenges that you have detected. If you want to share them with us, we will be happy to explore how we can help.
Thank you for looking at the challenge and sharing your feedback.
There’s a mix of use cases that Surface can support. The concept for this challenge was based on gaps in Surface coverage I noticed on ways in the US and feedback from discussions with communities in OSM Africa. In those discussions, Surface is helpful for map visualization and is sometimes used to determine if off-road vehicles are needed to access remote villages.
As this is my first challenge in the US to improve OSM, I wanted to avoid going too big with a huge number of tasks for the whole country and all highway classifications. Starting out with motorways seemed like a decent option, as the scope is limited, and there’s generally good source material coverage (aerial imagery, Bing Streetside, and Mapillary.)
Just curious why TomTom is so concerned about highway=motorway surface values. I’m confident that all of California’s motorway sections are paved. Or do we care about the difference between concrete and asphalt for some reason?
Goodhart’s Law, maybe? We’ve seen similar things elsewhere (not from TomTom), adding “roads” within private industrial areas that aren’t a general benefit to routing.
I share some of the opinions voiced in the mailing list thread. Determining surface types purely from aerial imagery can be quite difficult. Street level imagery (Bing streetside or Mapillary) or a real world survey/knowledge is often necessary to make a correct determination. When a mapper is intimately familiar with an area they may come to know from experience what the common surface types the highway departments use and what they look on aerial imagery. Such a mapper may be able to make an educated guess from aerial imagery alone, but that level of knowledge should not be assumed for a Map Roulette challenge.
Any challenge or other directed mapping activity like this should have instructions stating the importance of consulting street level imagery or real world surveying and not relying only on aerial imagery. Another good thing would be to instruct mappers that if they do decide to make an educated guess based only on aerial imagery they should add source:surface=aerial imagery (or similar). This would make it clear to future mappers that the surface value could use a second check from a better source to confirm or replace it.
Adding surface values to roads is certainly beneficial, but of all the highway classes, it is least beneficial on motorways. If surface is not present on a motorway data consumers can safely assume surface=paved and be correct in most if not all cases. No one has presented an example of an unpaved motorway yet, and while I won’t say with 100% certainty that every single motorway in the whole world is paved, it seems most likely. There is nothing wrong with adding surface values to motorways, but it would be a far greater impact to focus on any of the lower classes of highways. In rural parts of the United States, it is quite common to find unpaved residential, unclassified, tertiary, and possibly even secondary roads. Focusing on any or all of these would be a great Map Roulette challenge and would be far more beneficial than motorways.
I don’t sense a consensus yet on the approach of using Paved versus a specific surface value
The consensus is that tagging surface=paved or surface=unpaved is a good first step and better than no surface tag, but replacing with a more specific value is better still. Data consumers only interested in the paved/unpaved distinction should interpret asphalt, concrete, etc as paved and compacted, dirt, fine_gravel as unpaved. This is well documented on the surface wiki page.
+1 to Zeke’s comments.
It’s easy to mistag surface from imagery. In my opinion it’s better to have no surface tag than the wrong one. I just went back and looked at a local road that I fixed that was mis-tagged based on imagery as asphalt & it is fairly obvious that it is unpaved when you look at the intersection with the paved road. If there is a MapRoulette challenge it needs to be emphasized, don’t change a surface value that’s already there & be very careful & thorough using the imagery.
An easier remote task would be checking the validity of the highway=residential tags, there are many that should be unclassified or track.
If TomTom is interested in backroads, a good task would be verifying that the tiger imported roads on OSM are open to motorvehicles or private driveways.
IMHO by far the two most important tasks related to highways are:
Adding new streets in new subdivisions, especially high growth states like Florida where many new subdivisions are not represented yet.
Cleaning up / clearing out unedited or sparsely edited original TIGER that is still showing thousands of miles of “residential” streets roaming through forests and ag lands all over the USA.
Paved vs unpaved for routes regularly traveled by cars seems important and useful, especially for biking. Maybe its region-specific how much the surface type matters, but doesn’t seem important given the problems with missing new roads and flat out wrong info with unedited tiger in rural / uninhabited areas.
Personally, I’d put that way above adding surface tags to roads in importance, at least in California. Like @ezekielf said, on a motorway it’s safe to assume surface=paved is correct in most or all cases. So there isn’t that much benefit to tagging them that way in a lot of cases. The thousands of miles of residential roads in forests and ag land is a real problem though. One that it doesn’t seem there’s been much movement on. At least not in far Northern California where mainly do my mapping. Going straight to adding surface tags without cleaning up the non-exiting roads kind of seems like putting the cart before the horse though. I’d be interested to know how many roads that don’t exist have surface tags now thanks to the MapRoulette challenge or just people doing semi-automated edits now that the surface is rendered in the main style.
For routing purpose mainly it’s interesting to map the difference from the expected value. Usually more or less highway=service and upwards are considered as surface=paved, highway=track and below as surface=unpaved. So it’s far more important to tag the surface=unpaved on a highway=tertiary than a surface=asphalt on a highway=residential.
Though if you look more detailed into routing, like considering average speeds for biking or routing for skaters, suddenly the surface and even smoothness getting kind of more important than the highway-definition. But on this level of detail you need local mappers interested in that data and their maintenance. Considering the TIGER-progress in most areas of the US this is far, far away form being realistic.
@Mapiate So for me rather than such a challenge some AI-based aerial image reader, detecting potential “critical” roads (default might be paved, but reality is unpaved or vice-versa) would be a better input for such a challenge.