[Talk-us] OSM Elevation Data

[Greg Troxel]

Mike N <niceman at att.net> writes:

>   There are at least 2 problem use cases with an an external elevation
> data set which is acquired from above such as from a plane or space
> shuttle.

I'll interpret that as a blend of "problems with using an elevation data
set which contains the height of the ground by location" and "problems
with using a data set that has outright errors or is very coarse".

I think routers need to use quality data first, before we address any
residual issues.  SRTM is just not quality data any more, even though it
was surely great in its day.

>  1. Roads through tunnels are assigned the elevation profile of the
> mountain above.   However in practice 99% of tunnels have a constant
> slope from one end to the other and a router could connect them with a
> straight line elevation between the 2 entrances.

Good point.  Arguably routers should be doing that.

The same issue may exist with bridges, depend on resolution and whether
the LIDAR data is raw or "bare earth" where structures are removed.
What is the elevation where there's a bridge?  I'd say it's the surface
under.  But same point about typical even slope.

>  2. Roads within canyons can have elevation errors - the road can be
> intermittently hidden behind the canyon wall due to the viewpoint of
> the measuring device.   The automatic profile as calculated has wild
> swings in elevation every time the road enters or leaves view.

I think you're arguing that some elevation data sets have bad data.
It's an interesting question how much OSM should work around that.

There's a related issue where there is elevation data that doesn't have
look angle bugs (perhaps because it's LIDAR from multiple flight lines),
but has limited spatial resolution, perhaps 100m grid.   There,
interpolation might be off.

>   #2  has no automatic solution.   One possibility would be to assign
> elevation to those sections of roads in OSM that are likely to have
> this type of error.   Are there other ways around creating OSM
> elevation tag(s) and the can of worms that would open?

I think the first step is to get routers to use the highest-quality data
that exists.  In the US, it doesn't make senes to use SRTM data.   I
realize it varies, but in MA we have 1x1m, 1m vertical LIDAR data from
2010-2015 (multiple projects) covering the state.   MassGIS is
threatening to release new data (flown spring 2021) soon, which I think
might be 0.5x0.5m, 0.1m vertical.

I suspect that with good data, and tunnel/bridge processing the
remaining problems will be minor, and that it would be good to solve
those issues first and then discuss anew where we are.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 194 bytes
Desc: not available
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20221007/af9bf69d/attachment.sig>