[PROJ] Explaining geodesic distance discrepancies

[Charles Karney]

The ratio utm distance/geodesic distance is the "scale" of the UTM
transformation.  (The "tilting" you refer to is called the "meridian
convergence"; this doesn't affect the distance calculation because TM is
a conformal projection.)  For the midpoint of the geodesic you are
examining, the scale is 1.00028568317.  And the ratio you get is

   utm/geod = 1161.43346448816/1161.1017567084384 = 1.00028568364

So your utm and geod results are consistent.  I can't reproduce your
geod cmdline result.  If I run your python script, I get

     one: 17
     diff utm:          1161.433464475265
     diff geod:         1161.1017567084402
     diff geod cmdline: 1161.102

Perhaps you have the "wrong" geod in your path?

   --Charles

On 5/15/20 4:12 PM, Dima Kogan wrote:
> Hi. I'm seeing behavior I don't understand. Any clarification would be
> greatly appreciated.
> 
> I have two positions represented as latitude,longitude pairs, and I'm
> trying to get the distance between them. I've 3 methods that all yield
> different results, where I think they should be VERY close. The methods
> are:
> 
> - Converting the lat/lon to a Cartesian UTM system, and then using the
>    Cartesian distance
> 
> - pyproj.Geod
> 
> - The "geod" command-line tool
> 
> For an arbitrary pair of points that are ~1km apart, I see:
> 
> diff utm:          1161.43346448816
> diff geod:         1161.1017567084384
> diff geod cmdline: 1163.971
> 
> This looks wrong. The UTM probably should be slightly off because the
> two Cartesian systems are slightly tilted, but is 30cm over 1km really
> how far off it should be?
> 
> The difference between pyproj.Geod and the "geod" cmdline tool makes
> even less sense. In theory they're doing the same thing, so the results
> should be identical.
> 
> I'm attaching a Python script to run the 3 methods, and print the
> ranges.
> 
> Thanks!
> 
> 
> _______________________________________________
> PROJ mailing list
> PROJ at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/proj
>