[Tagging] RFC ele:regional

[Greg Troxel]

Martin Koppenhoefer <dieterdreist at gmail.com> writes:

> I was not aware there weren't any meaningful differences (when comparing
> some official height references to the German DHHN92 those in wikipedia.de
> with delta information all are within 1m besides Belgium DNG/TAW, which is
> -2.3).

Thanks for looking into this.  It is interesting about Belgium's datum
being so far off, but I'm sure there's some long-ago historical reason
and various datums along the way matched some earlier datum.   Perhaps
relating a nautical datum (tends to be mean low water) with a land datum
(tends to be mean sea level).

> Maybe all we should do is clarify that we are NOT expecting
> ellipsoid heights in ele tags (leaving open the possibility to add
> ele:datum tags)?

I would very much like to clarify that, and it feels like we have
consensus on that point.  I can edit the wiki page to make it clearer,
since we have had a big discussion and there is no one advocating for
ellipsoidal height in OSM.

I am ok with some text about the possible future use of ele:datum to
refer to some other datum, although I think it's preferable for people
to transform, just as it is for horizontal.

> WGS84 uses a 2 dimensional ellipsoidal coordinate system?

WGS84 at its core is a 3-dimensional cartesian system, written XYZ, with
the origin at the center of mass of the earth.  There is a transform to
latitude, longitude, and ellipsoidal height (which is just math and not
a source of errors).   All modern datums are like this.  Note that WGS84
is essentially the same thing as ITRF2014, the current international
datum from the geodesy (since of measuring the earth) community, perhaps
differing by a few mm, perhaps not.

> Wouldn't "natural" height information be based on this ellipsoid? I'm all
> fine with stating we should use geoid heights, but it doesn't necessarily
> seem to be implicit.

It does need to be made clear, as there is ample reason to think there
has been confusion about it in OSM.


ellipsoidal heights are not natural!

To understand this, one has to look to the history of surveying.  Until
the satellite era, horizontal datums and vertical datums were basically
separate.  Horizontal was done by maesured baselines and triangulation.

Vertical was done by 'leveling' which is basically a telescope with a
bubble level, so you transfer horizontally from one place to another,
and then read the difference on a rod, which is basically a giant
measuring stick.  As you move from a tide gauge across a continent, you
keep track of the accumulated differences (and loops, and then you do
least squares).

Implicit in leveling is that two places at the same height are at the
same gravitational potential, and that when you say "height is x m" you
mean that if you measured vertically (along the plumb line) it was x
until you got to where the ocean would flow to if there were a tunnel.
So height as surveyors used it, and the national/regional datums taht
this height is referenced too, have always been about gravity, and the 0
reference level has more or less been at some notion of sea level.
Typical is the mean reading of a tide gauge over 19 years (a sun/moon
cycle).

NAVD88 is referenced to a single tide gauge at Father's Point in
Rimouski, Quebec.  NGVD29 set a number of tide gauges around the
continent at 0 (and this caused distortions).

Satellite techniques just measure position and distance, and not
gravity.  Thus they in their raw form give you ellipsoidal height.  But,
this is not useful for water engineering, because lower elllipsoidal
height is not downhill.  So WGS84 defines a "geoid model" that gives the
height of an equal-gravity surface that more or less corresponds to mean
sea level, compared to the ellipsoid, and except for 1) some buggy
android programs and 2) survey-grade GPS equipment, all GPS (GNSS,
GLONASS, Galileo, BeiDou, etc.) receivers give altitudes in height above
the geoid, because aligning (at the meter level) with the previous
national height datums is important, and having water flow to lower
elevations is important.  And because nobody except national-level
surveyors has ever cared about ellipsoidal heights.

Ellipsoidal heights are basically used only as internal steps in
surveying, and nobody in the surveying world would ever think to put
them on a sign, or say to the public "the ellipsoidal height of Mount
Washington is X".  (A datasheet for surveyors would say that, but it
would also give the NAVD88 height.  A poll of 1000 surveyors asking
"should we put ellipsoidal height or NAVD88 on the sign" would result in
all 1000 of them looking at you like you were crazy and saying "NAVD88,
of course".)