[Talk-us-massachusetts] Braintree buildings+(Datum shift)

[Greg Troxel]

"Wayne Emerson, Jr. via Talk-us-massachusetts"
<talk-us-massachusetts at openstreetmap.org> writes:

> I first noticed this shift about a year ago when the iD editor
> software switched to using a different TMS server for the L3 parcel
> map. I later saw this email talked about the change of tile servers:

I saw this too in josm.

> https://lists.openstreetmap.org/pipermail/talk-us-massachusetts/2018-November/000424.html 
>
> At the time I was not subscribed to this list so I posted about it here:
>
> https://forum.openstreetmap.org/viewtopic.php?id=64699
>
> Not sure if my conclusions there were correct however.

I think you got it right but missing some signficant
complications.

> I had assumed that the downloadable files from MassGIS stayed fixed at
> the original datum, while the TMS applied some sort of techtonic plate
> movement based correction. The yearly divergence rate noted in my
> forum post is not big enough to account for the difference we see
> today, and may be partly due to the initial error in the original
> datum. Probably only MassGIS would know the source(s) of the
> difference.

(I don't see this notion or yearly divergence in the forum post.)


This is a really complicated subject, and I'll try to summarize without
being inaccurate.

Complication 1:

  Both NAD83 and WGS84 have had multiple realizations over the years,
  and those realizations are increasingly accurate.

Complication 2:

  Even coordinates of stations on our plate in NAD83 change over time.

The current version of NAD83 is called NAD83(2011).  More or less, the
intent is that points that are fixed to the crust on the North American
Plate (all of Mass, and most of the continental US/Canada -- except some
bits of the west coast) are intended to have stable coordinates.  If you
look closely, I think stations will have velocities in NAD83(2011) that
are about 2 mm/year.

The current version of WGS84 is WGS84(G1762), which is basically the
same as ITRF2008.  This intends to more or less have station coordinates
be stable, but in an average global way rather than a plate-fixed way.
In New England, the same stations that move 2 mm/year in NAD83(2011)
have velocities of maybe 15 mm/year.

Complication 3:

  WGS84 is a name for a group of datums, and when it is said unqualified
  like that, there is uncertainty at the 2m level because you don't know
  which version of WGS84 was used.  (This is also true for NAD83, but
  less so.)  This causes programs asked to transform from NAD83 to WGS84
  to say "they are the same, to within the fuzz they already each have".
  But if you ask to transform from NAD83(2011) to WGS84(G1762), they are
  not the same.

Complication 4:

  (slightly unsure here) TMS is defined in terms of "web mercator" which
  ia defined to be "WGS84" which leads to uncertainty.  I think that
  "obviously" TMS should be viewed as being WGS84(G1762) today.  But I
  suspect that if you got a bunch of open-source geodesy types together,
  that would be a heated argument.

The whole notion of worrying about this difference is relatively modern;
in the 80s and early 90s NAD83 and WGS84 were usually treated as
equivalent.

So this leads to questions:

  When MassGIS publishes data and says it is in "NAD83", which
  realization do they mean?

  (When MassGIS offers to convert to "WGS84" when downloading from
  OLIVER, which realization do they mean, and how do they convert?)

  When the old L3 tiles were created, how was the data obtained and
  transformed?

  What datum is the current L3 parcels TMS in?  Did MassGIS transform
  from NAD83(?) to WGS84(G1762)?  Or just assume equivalence?


I have a contact at MassGIS and am going to ask about this, once I
figure out as much as I can myself.


>> (I see what you mean about Braintree.)
>
> Looks like mainly north half of Braintree is a mess. These imports
> were done by crschmidt in January 2009.

That's the super early batch, and I'm not surprised.  crschmidt is no
longer active, but he imported (back in the days before people got upset
about imports) a ton of MassGIS data which had the really great effect
of MA not getting Tiger.  Some of the data is a little wonky (these
buildings apparently, and the open space layer), but mostly I think it's
been very positive.

I started working on OSM just after that import, after a friend pointed
it out to me.  Before that, OSM didn't seem clearly viable and
afterwards it was amazing.

> These are some of the changesets I found on bad Braintree buildings
> but It looks like the changesets cover huge areas:
>
> https://www.openstreetmap.org/changeset/727079
>
> https://www.openstreetmap.org/changeset/743822
> <https://www.openstreetmap.org/changeset/727079>
>
> https://www.openstreetmap.org/changeset/743003
> <https://www.openstreetmap.org/changeset/727079>
>
> The later imports done in 2013 look good.

Those are the ones led by Jason (who did most of the data transformation
work, with shared QA), and we were very careful.  I'm glad those are ok.